Schachtman Law » statistical evidence

TORTINI

For your delectation and delight, desultory dicta on the law of delicts.

Practical Solutions for the Irreproducibility Crisis

March 3rd, 2020

I have previously praised the efforts of the National Association of Scholars (NAS) for its efforts to sponsor a conference on “Fixing Science: Practical Solutions for the Irreproducibility Crisis.” The conference was a remarkable event, with a good deal of diverse view points, civil discussion and debate, and collegiality.

The NAS has now posted a follow up to its conference, with a link to slide presentations, and to a You Tube page with videos of the presentations. The NAS, along with The Independent Institute, should be commended for their organizational efforts, and their transparency in making the conference contents available now to a wider audience.

The conference took place on February 7th and 8th, and I had the privilege of starting the event with my presentation, “Not Just an Academic Dispute: Irreproducible Scientific Evidence Renders Legal Judgments Unsafe”.

Some, but not all, of the interesting presentations that followed:

Tim Edgell, “Stylistic Bias, Selective Reporting, and Climate Science” (Feb. 7, 2020)

Patrick J. Michaels, “Biased Climate Science” (Feb. 7, 2020)

Daniele Fanelli, “Reproducibility Reforms if there is no Irreproducibility Crisis” (Feb. 8, 2020)

On Saturday, I had the additional privilege of moderating a panel on “Group Think” in science, and its potential for skewing research focus and publication:

Lee Jussim, “Intellectual Diversity Limits Groupthink in Scientific Psychology” (Feb. 8, 2020)

Mark Regnerus, “Groupthink in Sociology” (Feb. 8, 2020)

Michael Shermer, “Giving the Devil His Due” (Feb. 8, 2020)

Later on Saturday, the presenters turned to methodological issues, many of which are key to understanding ongoing scientific and legal controversies:

Stanley Young, “Prevention and Management of Acute and Late Toxicities in Radiation Oncology”

James E. Enstrom, “Reproducibility is Essential to Combating Environmental Lysenkoism”

Deborah Mayo, “P-Value ‘Reforms’: Fixing Science or Threats to Replication and Falsification?” (Feb. 8, 2020)

Ronald L. Wasserstein, “What Professional Organizations Can Do To Fix The Irreproducibility Crisis” (Feb. 8, 2020)

Louis Anthony Cox, Jr., “Causality, Reproducibility, and Scientific Generalization in Public Health” (Feb. 8, 2020)

David Trafimow, “What Journals Can Do To Fix The Irreproducibility Crisis” (Feb. 8, 2020)

David Randall, “Regulatory Science and the Irreproducibility Crisis” (Feb. 8, 2020)

Posted in Scientific Evidence, Scientific Misconduct, Scientific Publishing, statistical evidence | Comments Off on Practical Solutions for the Irreproducibility Crisis

Counter Cancel Culture – The NAS Conference on Irreproducibility

February 9th, 2020

“The meaning of the world is the separation of wish and fact.” Kurt Gödel

Back in October 2019, David Randall, the Director of Research, of the National Association of Scholars, contacted me to ask whether I would be interested in presenting at a conference, to be titled “Fixing Science: Practical Solutions for the Irreproducibility Crisis.” David explained that the conference would be aimed at a high level consideration of whether such a crisis existed, and if so, what salutary reforms might be implemented.

As for the character and commitments of the sponsoring organizations, David was candid and forthcoming. I will quote him, without his permission, and ask his forgiveness later:

“The National Association of Scholars is taken to be conservative by many scholars; the Independent Institute is (broadly speaking) in the libertarian camp. The NAS is open to but currently agnostic about the degree of human involvement in climate change. The Independent Institute I take to be institutionally skeptical of consensus climate change theory–e.g., they recently hosted Willie Soon for lecture. A certain number of speakers prefer not to participate in events hosted by institutions with these commitments.”

To me, the ask was for a presentation on how the so-called replication crisis, or the irreproducibility crisis, affected the law. This issue was certainly one I have had much occasion to consider. Although I am aware of the “adjacency” arguments made by some that people should be mindful of whom they align with, I felt that nothing in my participation would compromise my own views or unduly accredit institutional positions of the sponsors.

I was flattered by the invitation, but I did some due diligence on the sponsoring organizations. I vaguely recalled the Independent Institute from my more libertarian days, but the National Association of Scholars (NAS, not to be confused with Nathan A. Schachtman) was relatively unknown to me. A little bit of research showed that the NAS had a legitimate interest in the irreproducibility crisis. David Randall had written a monograph for the organization, which was a nice summary of some of the key problems. The Irreproducibility Crisis of Modern Science: Causes, Consequences,and the Road to Reform (2018).

On other issues, the NAS seemed to live up to its description as “an organization of scholars committed to higher education as the catalyst of American freedom.” I listened to some of the group’s podcasts, Curriculum Vitae, and browsed through its publications. I found myself agreeing with many positions articulated by or through the NAS, and disagreeing with a few positions very strongly.

In looking over the list of other invited speakers, I saw great diversity of view points and approaches, One distinguished speaker, Daniele Fanelli, had criticized the very notion that there was a reproducibility crisis. In the world of statistics, there were strong defenders of statistical tests, and vociferous critics. I decided to accept the invitation, not because I was flattered, but because the replication issue was important, and I believed that I could add something to the discussion before an audience of professional scientists, statisticians, and educated lay persons. In writing to David Randall to accept the invitation, I told him that with respect to the climate change issues, I was not at all put off by healthy skepticism in the face all dogmas. Every dogma will have its day.

I did not give any further consideration to the political aspect of the conference until early January, when I received an email from a scientist, Lenny Teytelman, Ph.D., the C.E.O. of a company protocols.io, which addresses reproducibility issues. Dr Teytelman’s interest in improving reproducibility seemed quite genuine, but he wrote to express his deep concern about the conference and the organizations that were sponsoring it.

Perhaps a bit pedantically, he cautioned me that the NAS was not the National Academy of Sciences, a confusion that never occurred to me because the National Academies has been known as the National Academies of Science, Engineering and Medicine for several years now. Dr. Teytelman’s real concern seemed to be that the NAS is a “‘politically conservative advocacy group’.” (The internal scare quotes were Teytelman’s, but I was not afraid.) According to Dr. Teytelman, the NAS sought to undermine climate science and environmental protection by advancing a call for more reproducible science. He pointed me to what he characterized as an exposé on NAS, in Undark,¹ and he cautioned me that the National Association of Scholars’ work is “dangerous.” Finally, Dr. Teytelman urged me to reconsider my decision to participate in the conference.

I did reconsider my decision, but reaffirmed it in an email I sent back to Dr. Teytelman. I realized that I could be wrong, in which case, I would eat my words, confident that they would be most digestible:

Dear Dr Teytelman,

Thank you for your note. I was aware of the piece on Undark’s website, as well as the difference between the NAS and the NASEM. I don’t believe anyone involved in science education would likely to be confused between the two organizations. A couple of years ago, I wrote a teaching module on biomedical causation for the National Academies. This is my first presentation at the request of the NAS, and frankly I am honored by the organization’s request that I present at its conference.

I have read other materials that have been critical of the NAS and its publications on climate change and other issues. I know that there are views of the organization from which I would dissent, but I do not see my disagreement on some issues as a reason not to attend, and present at a conference on an issue of great importance to the legal system.

I am hardly an expert on climate change issues, and that is my failing. Most of my professional work involves health effects regulation and litigation. If the NAS has advanced sophistical arguments against a scientific claim, then the proper antidote will be to demonstrate its fallacious reasoning and misleading marshaling of evidence. I should think, however, as someone interested in improving the reproducibility of scientific research, you will agree that there is much common ground for discussion and reform of scientific practice, on a broader arrange [sic] of issues than climate change.

As for the political ‘conservatism’, of the organization, I am not sure why that is a reason to eschew participation in a conference that should be of great importance to people of all political views. My own politics probably owe much to the influence of Michael Oakeshott, which puts me in perhaps the smallest political tribe of any in the United States. If conservatism means antipathy to post-modernism, identity politics, political orthodoxies, and assaults on Enlightenment values and the Rule of Law, then count me in.

In any event, thanks for your solicitude. I think I can participate and return with my soul intact.

All the best.

Nathan

To his credit, Dr. Teytelman tenaciously continued. He acknowledged that the political leanings of the organizers were not a reason to boycott, but he politely pressed his case. We were now on a first name basis:

Dear Nathan,

I very much applaud all efforts to improve the rigour of our science. The problem here is that this NAS organization has a specific goal – undermining the environmental protection and denying climate change. This is why 7 out of the 21 speakers at the event are climate change deniers. [https://docs.google.com/spreadsheets/d/136FNLtJzACc6_JbbOxjy2urbkDK7GefRZ/edit?usp=sharing] And this isn’t some small fringe effort to be ignored. Efforts of this organization and others like them have now gotten us to the brink of a regulatory change at the United States Environmental Protection Agency which can gut the entire EPA (see a recent editorial against this I co-authored). This conference is not a genuine effort to talk about reproducibility. The reproducibility part is a clever disguise for pushing a climate change denialism agenda.

Best,

Lenny

I looked more carefully at Lenny’s spreadsheet, and considered the issue afresh. We were both pretty stubborn:

Dear Lenny,

Thank you for this information. I will review with interest.

I do not see that the conference is primarily or even secondarily about climate change vel non. There are two scientists, Trafimow and Wasserstein with whom I have some disagreements about statistical methodology. Tony Cox and Stan Young, whatever their political commitments or views on climate change may be, are both very capable statisticians, from whom I have learned a great deal. The conference should be a lively conversation about reproducibility, not about climate change. Given your interests and background, you should go.

I believe that your efforts here are really quite illiberal, although they are in line with the ‘cancel culture’, so popular on campuses these days.

Forty three years ago, I entered a Roman Catholic Church to marry the woman I love. There were no lightning bolts or temblors, even though I was then and I am now an atheist. Yes, I am still married to my first wife. Although I share the late Christopher Hitchins’ low view of the Catholic Church, somehow I managed to overcome my antipathy to being married in what some would call a house of ill repute. I even manage to agree with some Papist opinions, although not for the superstitious ‘reasons’ Papists embrace.

If I could tolerate the RC Church’s dogma for a morning, perhaps you could put aside the dichotomous ‘us and them’ view of the world and participate in what promises to be an interesting conference on reproducibility?

All the best.

Nathan

Lenny kindly acknowledged my having considered his issues, and wrote back a nice note, which I will quote again in full without permission, but with the hope that he will forgive me and even acknowledge that I have given his views an airing in this forum.

Hi Nathan,

We’ll have to agree to disagree. I don’t want to give a veneer of legitimacy to an organization whose goal is not improving reproducibility but derailing EPA and climate science.

Warmly,

Lenny

The business of psychoanalyzing motives and disparaging speakers and conference organizers is a dangerous business for several reasons. First motives can be inscrutable. Second, they can be misinterpreted. And third, they can be mixed. When speaking of organizations, there is the further complication of discerning a corporate motive among the constituent members.

The conference was an exciting, intellectually challenging event, which took place in Oakland, California, on February 7 and 8. I can report back to Lenny that his characterizations of and fears about the conference were unwarranted. While there were some assertions of climate change skepticism made with little or no evidence, the evidence-based presentations essentially affirmed climate change and sought to understand its causes and future course in a scientific way. But climate change was not why I went to this conference. On the more general issue of reform of scientific procedures and methods, we had open debates, some agreement on important principles, and robust and reasoned disagreement.

Lenny, you were correct that the NAS should not be ignored, but you should have gone to the meeting and participated in the conversation.

1 Michael Schulson, “A Remedy for Broken Science, Or an Attempt to Undercut It?” Undark (April 18, 2018).

Posted in Cognitive Biases, Conflicts of Interest, Scientific Evidence, Scientific Misconduct, statistical evidence | Comments Off on Counter Cancel Culture – The NAS Conference on Irreproducibility

American Statistical Association – Consensus versus Personal Opinion

December 13th, 2019

Lawyers and judges pay close attention to standards, guidances, and consenus statements from respected and recognized professional organizations. Deviations from these standards may be presumptive evidence of malpractice or malfeasance in civil and criminal litigation, in regulatory matters, and in other contexts. One important, recurring situation arises when trial judges must act as gatekeepers of the admissibility of expert witness opinion testimony. In making this crucial judicial determination, judges will want to know whether a challenged expert witness has deviated from an accepted professional standard of care or practice.

In 2016, the American Statistical Association (ASA) published a consensus statement on p-values. The ASA statement grew out of a lengthy process that involved assembling experts of diverse viewpoints. In October 2015, the ASA convened a two-day meeting for 20 experts to meet and discuss areas of core agreement. Over the following three months, the participating experts and the ASA Board members continued their discussions, which led to the ASA Executive Committee’s approval of the statement that was published in March 2016.[1]

The ASA 2016 Statement spelled out six relatively uncontroversial principles of basic statistical practice.[2] Far from rejecting statistical significance, the six principles embraced statistical tests as an important but insufficient basis for scientific conclusions:

“3. Scientific conclusions and business or policy decisions should not be based only on whether a p-value passes a specific threshold.”

Despite the fairly clear and careful statement of principles, legal actors did not take long to misrepresent the ASA principles.[3] What had been a prescription about the insufficiency of p-value thresholds was distorted into strident assertions that statistical significance was unnecessary for scientific conclusions.

Three years after the ASA published its p-value consensus document, ASA Executive Director, Ronald Wasserstein, and two other statisticians, published an editorial in a supplemental issue of The American Statistician, in which they called for the abandonment of significance testing.[4] Although the Wasserstein’s editorial was clearly labeled as such, his essay introduced the special journal issue, and it appeared without disclaimer over his name, and his official status as the ASA Executive Director.

Sowing further confusion, the editorial made the following pronouncement:[5]

“The [2016] ASA Statement on P-Values and Statistical Significance stopped just short of recommending that declarations of ‘statistical significance’ be abandoned. We take that step here. We conclude, based on our review of the articles in this special issue and the broader literature, that it is time to stop using the term “statistically significant” entirely. Nor should variants such as ‘significantly different’, ‘p < 0.05’, and ‘nonsignificant’ survive, whether expressed in words, by asterisks in a table, or in some other way.”

The ASA is a collective body, and its ASA Statement 2016 was a statement from that body, which spoke after lengthy deliberation and debate. The language, quoted above, moves within one paragraph, from the ASA Statement to the royal “We,” who are taking the step of abandoning the term “statistically significant.” Given the unqualified use of the collective first person pronoun in the same paragraph that refers to the ASA, combined with Ronald Wasserstein’s official capacity, and the complete absence of a disclaimer that this pronouncement was simply a personal opinion, a reasonable reader could hardly avoid concluding that this pronouncement reflected ASA policy.

Your humble blogger, and others, read Wasserstein’s 2019 editorial as an ASA statement.[6] Although it is true that the 2019 paper is labeled “editorial,” and that the editorial does not describe a consensus process, there is no disclaimer such as is customary when someone in an official capacity publishes a personal opinion. Indeed, rather than the usual disclaimer, the Wasserstein editorial thanks the ASA Board of Directors “for generously and enthusiastically supporting the ‘p-values project’ since its inception in 2014.” This acknowledgement strongly suggests that the editorial is itself part of the “p-values project,” which is “enthusiastically” supported by the ASA Board of Directors.

If the editorial were not itself confusing enough, an unsigned email from “ASA <asamail@amstat.org>” was sent out in July 2019, in which the anonymous ASA author(s) takes credit for changing statistical guidelines at the New England Journal of Medicine:[7]

From: ASA <asamail@amstat.org>
Date: Thu, Jul 18, 2019 at 1:38 PM
Subject: Major Medical Journal Updates Statistical Policy in Response to ASA Statement
To: <XXXX>

The email is itself an ambiguous piece of evidence as to what the ASA is claiming. The email says that the New England Journal of Medicine changed its guidelines “in response to the ASA Statement on P-values and Statistical Significance and the subsequent The American Statistician special issue on statistical inference.” Of course, the “special issue” was not just Wasserstein’s editorial, but the 42 other papers. So this claim leaves open to doubt exactly what in the 2019 special issue the NEJM editors were responding to. Given that the 42 articles that followed Wasserstein’s editorial did not all agree with Wasserstein’s “steps taken,” or with each other, the only landmark in the special issue was the editorial over the name of the ASA’s Executive Director.

Moreover, a reading of the NEJM revised guidelines does not suggest that the journal’s editors were unduly influenced by the Wasserstein editorial or the 42 accompanying papers. The journal mostly responded to the ASA 2016 consensus paper by putting some teeth into its Principle 4, which dealt with multiplicity concerns in submitted manuscripts. The newly adopted (2019) NEJM author guidelines do not take step out with Wasserstein and colleagues; there is no general prohibition on p-values or statements of “statistical significance.”

The confusion propagated by the Wasserstein 2019 editorial has not escaped the attention of other ASA officials. An editorial in the June 2019 issue of AmStat News, by ASA President Karen Kafadar, noted the prevalent confusion and uneasiness over the 2019 The American Statistician special issue, the lack of consensus, and the need for healthy debate.[8]

In this month’s issue of AmStat News, President Kafadar returned to the issue of the confusion over the 2019 ASA special issue of The American Statistician, in her “President’s Corner.” Because Executive Director Wasserstein’s editorial language about “we now take this step” is almost certainly likely to find its way into opportunistic legal briefs, Kafadar’s comments are worth noting in some detail:[9]

“One final challenge, which I hope to address in my final month as ASA president, concerns issues of significance, multiplicity, and reproducibility. In 2016, the ASA published a statement that simply reiterated what p-values are and are not. It did not recommend specific approaches, other than ‘good statistical practice … principles of good study design and conduct, a variety of numerical and graphical summaries of data, understanding of the phenomenon under study, interpretation of results in context, complete reporting and proper logical and quantitative understanding of what data summaries mean’.

The guest editors of the March 2019 supplement to The American Statistician went further, writing: ‘The ASA Statement on P-Values and Statistical Significance stopped just short of recommending that declarations of “statistical significance” be abandoned. We take that step here. … [I]t is time to stop using the term “statistically significant” entirely’.

Many of you have written of instances in which authors and journal editors – and even some ASA members – have mistakenly assumed this editorial represented ASA policy. The mistake is understandable: The editorial was coauthored by an official of the ASA. In fact, the ASA does not endorse any article, by any author, in any journal – even an article written by a member of its own staff in a journal the ASA publishes.”

Kafadar’s caveat should quash incorrect assertions about the ASA’s position on statistical significance testing. It is a safe bet, however, that such assertions will appear in trial and appellate briefs.

Statistical reasoning is difficult enough for most people, but the hermeneutics of American Statistical Association publications on statistical significance may require a doctorate of divinity degree. In a cleverly titled post, Professor Deborah Mayo argues that there is no other way to interpret the Wasserstein 2019 editorial except as laying down an ASA prescription. Deborah G. Mayo, “Les stats, c’est moi,” Error Philosophy (Dec. 13, 2019). I accept President Kafadar’s correction at face value, and accept that I, like many other readers, misinterpreted the Wasserstein editorial as having the imprimatur of the ASA. Mayo points out, however, that Kafadar’s correction in a newsletter may be insufficient at this point, and that a stronger disclaimer is required. Officers of the ASA are certainly entitled to their opinions and the opportunity to present them, but disclaimers would bring clarity and transparency to published work of these officials.

Wasserstein’s 2019 editorial goes further to make a claim about how his “step” will ameliorate the replication crisis:

“In this world, where studies with ‘p < 0.05’ and studies with ‘p > 0.05 are not automatically in conflict, researchers will see their results more easily replicated – and, even when not, they will better understand why.”

The editorial here seems to be attempting to define replication failure out of existence. This claim, as stated, is problematic. A sophisticated practitioner may think of the situation in which two studies, one with p = .048, and another with p = 0.052 might be said not to be conflict. In real world litigation, however, advocates will take Wasserstein’s statement about studies not in conflict (despite p-values above and below a threshold, say 5%) to the extremes. We can anticipate claims that two similar studies with p-values above and below 5%, say with one p-value at 0.04, and the other at 0.40, will be described as not in conflict, with the second a replication of the first test. It is hard to see how this possible interpretation of Wasserstein’s editorial, although consistent with its language, will advance sound, replicable science.[10]

[1] Ronald L. Wasserstein & Nicole A. Lazar, “The ASA’s Statement on p-Values: Context, Process, and Purpose,” 70 The Am. Statistician 129 (2016).

[2] “The American Statistical Association’s Statement on and of Significance” (Mar. 17, 2016).

[3] See, e.g., “The Education of Judge Rufe – The Zoloft MDL” (April 9, 2016) (Zoloft litigation); “The ASA’s Statement on Statistical Significance – Buzzing from the Huckabees” (Mar. 19, 2016); “The American Statistical Association Statement on Significance Testing Goes to Court – Part I” (Nov. 13, 2018).

[4] Ronald L. Wasserstein, Allen L. Schirm, and Nicole A. Lazar, “Editorial: Moving to a World Beyond ‘p < 0.05’,” 73 Am. Statistician S1, S2 (2019).

[5] Id. at S2.

[6] See “Has the American Statistical Association Gone Post-Modern?” (Mar. 24, 2019); Deborah G. Mayo, “The 2019 ASA Guide to P-values and Statistical Significance: Don’t Say What You Don’t Mean,” Error Statistics Philosophy (June 17, 2019); B. Haig, “The ASA’s 2019 update on P-values and significance,” Error Statistics Philosophy (July 12, 2019).

[7] See “Statistical Significance at the New England Journal of Medicine” (July 19, 2019); See also Deborah G. Mayo, “The NEJM Issues New Guidelines on Statistical Reporting: Is the ASA P-Value Project Backfiring?” Error Statistics Philosophy (July 19, 2019).

[8] See Kafadar, “Statistics & Unintended Consequences,” AmStat News 3,4 (June 2019).

[9] Karen Kafadar, “The Year in Review … And More to Come,” AmStat News 3 (Dec. 2019).

[10] See also Deborah G. Mayo, “P‐value thresholds: Forfeit at your peril,” 49 Eur. J. Clin. Invest. e13170 (2019).

Posted in statistical evidence | Comments Off on American Statistical Association – Consensus versus Personal Opinion

Is the IARC Lost in the Weeds?

November 30th, 2019

A couple of years ago, I met David Zaruk at a Society for Risk Analysis meeting, where we were both presenting. I was aware of David’s blogging and investigative journalism, but meeting him gave me a greater appreciation for the breadth and depth of his work. For those of you who do not know David, he is present in cyberspace as the Risk-Monger who blogs about risk and science communications issues. His blog has featured cutting-edge exposés about the distortions in risk communications perpetuated by the advocacy of non-governmental organizations (NGOs). Previously, I have recorded my objections to the intellectual arrogance of some such organizations that purport to speak on behalf of the public interest, when often they act in cahoots with the lawsuit industry in the manufacturing of tort and environmental litigation.

David’s writing on the lobbying and control of NGOs by plaintiffs’ lawyers from the United States should be required reading for everyone who wants to understand how litigation sausage is made. His series, “SlimeGate” details the interplay among NGO lobbying, lawsuit industry maneuvering, and carcinogen determinations at the International Agency for Research on Cancer (IARC). The IARC, a branch of the World Health Organization, is headquartered in Lyon, France. The IARC convenes “working groups” to review the scientific studies of the carcinogencity of various substances and processes. The IARC working groups produce “monographs” of their reviews, and the IARC publishes these monographs, in print and on-line. The United States is in the top tier of participating countries for funding the IARC.

The IARC was founded in 1965, when observational epidemiology was still very much an emerging science, with expertise concentrated in only a few countries. For its first few decades, the IARC enjoyed a good reputation, and its monographs were considered definitive reviews, especially under its first director, Dr. John Higginson, from 1966 to 1981.[1] By the end of the 20^th century, the need for the IARC and its reviews had waned, as the methods of systematic review and meta-analyses had evolved significantly, and had became more widely standardized and practiced.

Understandably, the IARC has been concerned that the members of its working groups should be viewed as disinterested scientists. Unfortunately, this concern has been translated into an asymmetrical standard that excludes anyone with a hint of manufacturing connection, but keeps the door open for those scientists with deep lawsuit industry connections. Speaking on behalf of the plaintiffs’ bar, Michael Papantonio, a plaintiffs’ lawyer who founded Mass Torts Made Perfect, noted that “We [the lawsuit industry] operate just like any other industry.”[2]

David Zaruk has shown how this asymmetry has been exploited mercilessly by the lawsuit industry and its agents in connection with the IARC’s review of glyphosate.[3] The resulting IARC classification of glyphosate has led to a litigation firestorm and an all-out assault on agricultural sustainability and productivity.[4]

The anomaly of the IARC’s glyphosate classification has been noted by scientists as well. Dr. Geoffrey Kabat is a cancer epidemiologist, who has written perceptively on the misunderstandings and distortions of cancer risk assessments in various settings.[5] He has previously written about glyphosate in Forbes and elsewhere, but recently he has written an important essay on glyphosate in Issues in Science and Technology, which is published by the National Academies of Sciences, Engineering, and Medicine and Arizona State University. In his essay, Dr. Kabat details how the IARC’s evaluation of glyphosate is an outlier in the scientific and regulatory world, and is not well supported by the available evidence.[6]

The problems with the IARC are both substantive and procedural.[7] One of the key problems that face IARC evaluations is an incoherent classification scheme. IARC evaluations classify putative human carcinogenic risks into five categories: Group I (known), Group 2A (probably), Group 2B (possibly), Group 3 (unclassifiable), and Group 4 (probably not). Group 4 is virtually an empty set with only one substance, caprolactam ((CH₂)₅C(O)NH), an organic compound used in the manufacture of nylon.

In the IARC evaluation at issue, glyphosate was placed into Group 2A, which would seem to satisfy the legal system’s requirement that an exposure more likely than not causes the harm in question. Appearances and word usage, however, can be deceiving. Probability is a continuous scale from zero to one. In Bayesian decision making, zero and one are unavailable because if either was our starting point, no amount of evidence could ever change our judgment of the probability of causation. (Cromwell’s Rule) The IARC informs us that its use of “probably” is quite idiosyncratic; the probability that a Group 2A agent causes cancer has “no quantitative” meaning. All the IARC intends is that a Group 2A classification “signifies a greater strength of evidence than possibly carcinogenic.”[8]

In other words, Group 2A classifications are consistent with having posterior probabilities of less than 0.5 (or 50 percent). A working group could judge the probability of a substance or a process to be carcinogenic to humans to be greater than zero, but no more than five or ten percent, and still vote for a 2A classification, in keeping with the IARC Preamble. This low probability threshold for a 2A classification converts the judgment of “probably carcinogenic” into a precautionary prescription, rendered when the most probable assessment is either ignorance or lack of causality. There is thus a practical certainty, close to 100%, that a 2A classification will confuse judges and juries, as well as the scientific community.

In IARC-speak, a 2A “probability” connotes “sufficient evidence” in experimental animals, and “limited evidence” in humans. A substance can receive a 2A classification even when the sufficient evidence of carcinogenicity occurs in one non-human animal specie, even though other animal species fail to show carcinogenicity. A 2A classification can raise the thorny question in court whether a claimant is more like a rat or a mouse.

Similarly, “limited evidence” in humans can be based upon inconsistent observational studies that fail to measure and adjust for known and potential confounding risk factors and systematic biases. The 2A classification requires little substantively or semantically, and many 2A classifications leave juries and judges to determine whether a chemical or medication caused a human being’s cancer, when the basic predicates for Sir Austin Bradford Hill’s factors for causal judgment have not been met.[9]

In courtrooms, IARC 2A classifications should be excluded as legally irrelevant, under Rule 403. Even if a 2A IARC classification were a credible judgment of causation, admitting evidence of the classification would be “substantially outweighed by a danger of … unfair prejudice, confusing the issues, [and] misleading the jury….”[10]

The IARC may be lost in the weeds, but there is no need to fret. A little Round Up™ will help.

[1] See John Higginson, “The International Agency for Research on Cancer: A Brief History of Its History, Mission, and Program,” 43 Toxicological Sci. 79 (1998).

[2] Sara Randazzo & Jacob Bunge, “Inside the Mass-Tort Machine That Powers Thousands of Roundup Lawsuits,” Wall St. J. (Nov. 25, 2019).

[3] David Zaruk, “The Corruption of IARC,” Risk Monger (Aug. 24, 2019); David Zaruk, “Greed, Lies and Glyphosate: The Portier Papers,” Risk Monger (Oct. 13, 2017).

[4] Ted Williams, “Roundup Hysteria,” Slate Magazine (Oct. 14, 2019).

[5] See, e.g., Geoffrey Kabat, Hyping Health Risks: Environmental Hazards in Everyday Life and the Science of Epidemiology (2008); Geoffrey Kabat, Getting Risk Right: Understanding the Science of Elusive Health Risks (2016).

[6] Geoffrey Kabat, “Who’s Afraid of Roundup?” 36 Issues in Science and Technology (Fall 2019).

[7] See Schachtman, “Infante-lizing the IARC” (May 13, 2018); “The IARC Process is Broken” (May 4, 2016). See also Eric Lasker and John Kalas, “Engaging with International Carcinogen Evaluations,” Law360 (Nov. 14, 2019).

[8] “IARC Preamble to the IARC Monographs on the Identification of Carcinogenic Hazards to Humans,” at Sec. B.5., p.31 (Jan. 2019); See also “IARC Advisory Group Report on Preamble” (Sept. 2019).

[9] See Austin Bradford Hill, “The Environment and Disease: Association or Causation?” 58 Proc. Royal Soc’y Med. 295 (1965) (noting that only when “[o]ur observations reveal an association between two variables, perfectly clear-cut and beyond what we would care to attribute to the play of chance,” do we move on to consider the nine articulated factors for determining whether an association is causal.

[10] Fed. R. Evid. 403.

Posted in Causation, Rule 702, statistical evidence | Comments Off on Is the IARC Lost in the Weeds?

Does the California State Bar Discriminate Unlawfully?

November 24th, 2019

Earlier this month, various news outlets announced a finding in a California study that black male attorneys are three times more likely to be disciplined by the State Bar than their white male counterparts.[1] Some of the news accounts treated the study findings as conclusions that the Bar had engaged in race discrimination. One particularly irresponsible website proclaimed that “bar discipline is totally racist.”[2] Indeed, the California State Bar itself apparently plans to hire consulting experts to help it achieve “bias-free decision-making and processes,” to eliminate “unintended bias,” and to consider how, if at all, to weigh prior complaints in the disciplinary procedure.[3]

The California Bar’s report was prepared by a social scientist, George Farkas, of the School of Education at University of California, Irvine. Based upon data from attorneys admitted to the California bar between 1990 and 2008, Professor Farkas reported crude prevalence rates of discipline, probation, disbarment, or resignation, by race.[4] The disbarment/ resignation rate for black male lawyers was 3.9%, whereas the rate for white male lawyers was 1%. Disparities, however, are not unlawful discriminations.

The disbarment/resignation rate for black female lawyers was 0.9%, but no one has suggested that there is implicit bias in favor of black women over both black and white male lawyers. White women were twice as likely as Asian women to resign, or be placed on probation or be disbarred (0.4% versus 0.2%).

The ABA’s coverage sheepishly admitted that “[d]ifferences could be explained by the number of complaints received about an attorney, the number of investigations opened, the percentage of investigations in which a lawyer was not represented by counsel, and previous discipline history.”[5]

Farkas’s report of October 31, 2019, was transmitted to the Bar’s Board of Trustees, on November 14^th.[6] As anyone familiar with discrimination law would have expected, Professor Farkas conducted multiple regression analyses that adjusted for the number of previous complaints filed against the errant lawyer, and whether the lawyer was represented by counsel before the Bar. The full analyses showed that these other important variables, not race – not could – but did explain variability in discipline rates:

“Statistically, these variables explained all of the differences in probation and disbarment rates by race/ethnicity. Among all variables included in the final analysis, prior discipline history was found to have the strongest effects [sic] on discipline outcomes, followed by the proportion of investigations in which the attorney under investigation was represented by counsel, and the number of investigations.”[7]

The number of previous complaints against a particular lawyer surely has a role in considering whether a miscreant lawyer should be placed on probation, or subjected to disbarment. And without further refinement of the analysis, and irrespective of race or ethnicity, failure to retain counsel for disciplinary hearings may correlate strongly with futility of any defense.

Curiously, the Farkas report did not take into account the race or ethnicity of the complainants before the Bar’s disciplinary committee. The Farkas report seems reasonable as far as it goes, but the wild conclusions drawn in the media would not pass Rule 702 gatekeeping.

[1] See, e.g., Emma Cueto, “Black Male Attorneys Disciplined More Often, California Study Finds,” Law360 (Nov. 18, 2019); Debra Cassens Weiss, “New California bar study finds racial disparities in lawyer discipline,” Am. Bar Ass’n J. (Nov. 18, 2019).

[2] Joe Patrice, “Study Finds That Bar Discipline Is Totally Racist Shocking Absolutely No One: Black male attorneys are more likely to be disciplined than white attorneys,” Above the Law (Nov. 19, 2019).

[3] Debra Cassens Weiss, “New California bar study finds racial disparities in lawyer discipline,” Am. Bar Ass’n J. (Nov. 18, 2019).

[4] George Farkas, “Discrepancies by Race and Gender in Attorney Discipline by the State Bar of California: An Empirical Analysis” (Oct. 31, 2019).

[5] Debra Cassens Weiss, supra at note 3.

[6] Dag MacLeod (Chief of Mission Advancement & Accountability Division) & Ron Pi (Principal Analyst, Office of Research & Institutional Accountability), Report on Disparities in the Discipline System (Nov. 14, 2019).

[7] Dag MacLeod & Pi, Report on Disparities in the Discipline System at 4 (Nov. 14, 2019) (emphasis added).

Posted in statistical evidence | Comments Off on Does the California State Bar Discriminate Unlawfully?

Palavering About P-Values

August 17th, 2019

The American Statistical Association’s most recent confused and confusing communication about statistical significance testing has given rise to great mischief in the world of science and science publishing.[1] Take for instance last week’s opinion piece about “Is It Time to Ban the P Value?” Please.

Helena Chmura Kraemer is an accomplished professor of statistics at Stanford University. This week the Journal of the American Medical Association network flagged Professor Kraemer’s opinion piece on p-values as one of its most read articles. Kraemer’s eye-catching title creates the impression that the p-value is unnecessary and inimical to valid inference.[2]

Remarkably, Kraemer’s article commits the very mistake that the ASA set out to correct back in 2016,[3] by conflating the probability of the data under a hypothesis of no association with the probability of a hypothesis given the data:

“If P value is less than .05, that indicates that the study evidence was good enough to support that hypothesis beyond reasonable doubt, in cases in which the P value .05 reflects the current consensus standard for what is reasonable.”

The ASA tried to break the bad habit of scientists’ interpreting p-values as allowing us to assign posterior probabilities, such as beyond a reasonable doubt, to hypotheses, but obviously to no avail.

Kraemer also ignores the ASA 2016 Statement’s teaching of what the p-value is not and cannot do, by claiming that p-values are determined by non-random error probabilities such as:

“the reliability and sensitivity of the measures used, the quality of the design and analytic procedures, the fidelity to the research protocol, and in general, the quality of the research.”

Kraemer provides errant advice and counsel by insisting that “[a] non-significant result indicates that the study has failed, not that the hypothesis has failed.” If the p-value is the measure of the probability of observing an association at least as large as obtained given an assumed null hypothesis, then of course a large p-value cannot speak to the failure of the hypothesis, but why declare that the study has failed? The study was perhaps indeterminate, but it still yielded information that perhaps can be combined with other data, or help guide future studies.

Perhaps in her most misleading advice, Kraemer asserts that:

“[w]hether P values are banned matters little. All readers (reviewers, patients, clinicians, policy makers, and researchers) can just ignore P values and focus on the quality of research studies and effect sizes to guide decision-making.”

Really? If a high quality study finds an “effect size” of interest, we can now ignore random error?

The ASA 2016 Statement, with its “six principles,” has provoked some deliberate or ill-informed distortions in American judicial proceedings, but Kraemer’s editorial creates idiosyncratic meanings for p-values. Even the 2019 ASA “post-modernism” does not advocate ignoring random error and p-values, as opposed to proscribing dichotomous characterization of results as “statistically significant,” or not.[4] The current author guidelines for articles submitted to the Journals of the American Medical Association clearly reject this new-fangled rejection of evaluating this new-fangled rejection of the need to assess the role of random error.[5]

[1] See Ronald L. Wasserstein, Allen L. Schirm, and Nicole A. Lazar, “Editorial: Moving to a World Beyond ‘p < 0.05’,” 73 Am. Statistician S1, S2 (2019).

[2] Helena Chmura Kraemer, “Is It Time to Ban the P Value?” J. Am. Med. Ass’n Psych. (August 7, 2019), in-press at doi:10.1001/jamapsychiatry.2019.1965.

[3] Ronald L. Wasserstein & Nicole A. Lazar, “The ASA’s Statement on p-Values: Context, Process, and Purpose,” 70 The American Statistician 129 (2016).

[4] “Has the American Statistical Association Gone Post-Modern?” (May 24, 2019).

[5] See instructions for authors at https://jamanetwork.com/journals/jama/pages/instructions-for-authors

Posted in statistical evidence, Uncategorized | Comments Off on Palavering About P-Values

Statistical Significance at the New England Journal of Medicine

July 19th, 2019

Some wild stuff has been going on in the world of statistics, at the American Statistical Association, and elsewhere. A very few obscure journals have declared p-values to be verboten, and presumably confidence intervals as well. The world of biomedical research has generally reacted more sanely, with authors defending the existing frequentist approaches and standards.[1]

This week, the editors of the New England Journal of Medicine have issued new statistical guidelines for authors. The Journal’s approach seems appropriately careful and conservative for the world of biomedical research. In an editorial introducing the new guidelines,[2] the Journal editors remind their potential authors that statistical significance and p-values are here to stay:

“Despite the difficulties they pose, P values continue to have an important role in medical research, and we do not believe that P values and significance tests should be eliminated altogether. A well-designed randomized or observational study will have a primary hypothesis and a prespecified method of analysis, and the significance level from that analysis is a reliable indicator of the extent to which the observed data contradict a null hypothesis of no association between an intervention or an exposure and a response. Clinicians and regulatory agencies must make decisions about which treatment to use or to allow to be marketed, and P values interpreted by reliably calculated thresholds subjected to appropriate adjustments have a role in those decisions.”[3]

The Journal’s editors described their revamped statistical policy as being based upon three premises:

(1) adhering to prespecified analysis plans if they exist;

(2) declaring associations or effects only for statistical analyses that have pre-specified “a method for controlling type I error”; and

(3) presenting evidence about clinical benefits or harms requires “both point estimates and their margins of error.”

With a hat tip to the ASA’s recent pronouncements on statistical significance,[4] the editors suggest that their new guidelines have moved away from bright-line applications of statistical significance “as a bright-line marker for a conclusion or a claim”[5]:

“[T]he notion that a treatment is effective for a particular outcome if P < 0.05 and ineffective if that threshold is not reached is a reductionist view of medicine that does not always reflect reality.”[6]

The editors’ language intimates greater latitude for authors in claiming associations or effects from their studies, but this latitude may well be circumscribed by tighter control over such claims in the inevitable context of multiple testing within a dataset.

The editors’ introduction of the new guidelines is not entirely coherent. The introductory editorial notes that the use of p-values for reporting multiple outcomes, without adjustments for multiplicity, inflates the number of findings with p-values less than 5%. The editors thus caution against “uncritical interpretation of multiple inferences,” which can be particularly threatening to valid inference when not all the comparisons conducted by the study investigators have been reported in their manuscript.[7] They reassuringly tell prospective authors that many methods are available to adjust for multiple comparisons, and can be used to control Type I error probability “when specified in the design of a study.”[8]

But what happens when such adjustment methods are not pre-specified in the study design? Failure to to do so do not appear to be disqualifying factors for publication in the Journal. For one thing, when the statistical analysis plan of the study has not specified adjustment methods for controlling type I error probabilities, then authors must replace p-values with “estimates of effects or association and 95% confidence intervals.”[9] It is hard to understand how this edict helps when the specified coefficient of 95% is a continuation of the 5% alpha, which would have been used in any event. The editors seem to be saying that if authors fail to pre-specify or even post-specify methods for controlling error probabilities, then they cannot declare statistical significance, or use p-values, but they can use confidence intervals in the same way they have been using them, and with the same misleading interpretations supplied by their readers.

More important, another price authors will have to pay for multiple testing without pre-specified methods of adjustment is that they will affirmatively have to announce their failure to adjust for multiplicity and that their putative associations “may not be reproducible.” Tepid as this concession is, it is better than previous practice, and perhaps it will become a badge of shame. The crucial question is whether judges, in exercising their gatekeeping responsibilities, will see these acknowledgements as disabling valid inferences from studies that carry this mandatory warning label.

The editors have not issued guidelines for the use of Bayesian statistical analyses, because “the large majority” of author manuscripts use only frequentist analyses.[10] The editors inform us that “[w]hen appropriate,” they will expand their guidelines to address Bayesian and other designs. Perhaps this expansion will be appropriate when Bayesian analysts establish a track record of abuse in their claiming of associations and effects.

The new guidelines themselves are not easy to find. The Journal has not published these guidelines as an article in their published issues, but has relegated them to a subsection of their website’s instructions to authors for new manuscripts:

https://www.nejm.org/author-center/new-manuscripts

Presumably, the actual author instructions control in any perceived discrepancy between this week’s editorial and the guidelines themselves. Authors are told that p-values generally should be two-sided. Authors’ use of:

“Significance tests should be accompanied by confidence intervals for estimated effect sizes, measures of association, or other parameters of interest. The confidence intervals should be adjusted to match any adjustment made to significance levels in the corresponding test.”

Similarly, the guidelines call for, but do not require, pre-specified methods of controlling family-wide error rates for multiple comparisons. For observational studies submitted without pre-specified methods of error control, the guidelines recommend the use of point estimates and 95% confidence intervals, with an explanation that the interval widths have not been adjusted for multiplicity, and a caveat that the inferences from these findings may not be reproducible. The guidelines recommend against using p-values for such results, but again, it is difficult to see why reporting the 95% confidence intervals is recommended when p-values are not recommended.

[1] Jonathan A. Cook, Dean A. Fergusson, Ian Ford, Mithat Gonen, Jonathan Kimmelman, Edward L. Korn, and Colin B. Begg, “There is still a place for significance testing in clinical trials,” 16 Clin. Trials 223 (2019).

[2] David Harrington, Ralph B. D’Agostino, Sr., Constantine Gatsonis, Joseph W. Hogan, David J. Hunter, Sharon-Lise T. Normand, Jeffrey M. Drazen, and Mary Beth Hamel, “New Guidelines for Statistical Reporting in the Journal,” 381 New Engl. J. Med. 285 (2019).

[3] Id. at 286.

[4] See id. (“Journal editors and statistical consultants have become increasingly concerned about the overuse and misinterpretation of significance testing and P values in the medical literature. Along with their strengths, P values are subject to inherent weaknesses, as summarized in recent publications from the American Statistical Association.”) (citing Ronald L. Wasserstein & Nicole A. Lazar, “The ASA’s statement on p-values: context, process, and purpose,” 70 Am. Stat. 129 (2016); Ronald L. Wasserstein, Allen L. Schirm, and Nicole A. Lazar, “Moving to a world beyond ‘p < 0.05’,” 73 Am. Stat. s1 (2019)).

[5] Id. at 285.

[6] Id. at 285-86.

[7] Id. at 285.

[8] Id., citing Alex Dmitrienko, Frank Bretz, Ajit C. Tamhane, Multiple testing problems in pharmaceutical statistics (2009); Alex Dmitrienko & Ralph B. D’Agostino, Sr., “Multiplicity considerations in clinical trials,” 378 New Engl. J. Med. 2115 (2018).

[9] Id.

[10] Id. at 286.

Posted in Scientific Evidence, statistical evidence | Comments Off on Statistical Significance at the New England Journal of Medicine

Science Bench Book for Judges

July 13th, 2019

On July 1^st of this year, the National Judicial College and the Justice Speakers Institute, LLC released an online publication of the Science Bench Book for Judges [Bench Book]. The Bench Book sets out to cover much of the substantive material already covered by the Federal Judicial Center’s Reference Manual:

Acknowledgments

Table of Contents

Introduction: Why This Bench Book?

What is Science?

Scientific Evidence

Introduction to Research Terminology and Concepts

Pre-Trial Civil

Pre-trial Criminal

Trial

Juvenile Court

The Expert Witness

Evidence-Based Sentencing

Post Sentencing Supervision

Civil Post Trial Proceedings

Conclusion: Judges—The Gatekeepers of Scientific Evidence

Appendix 1 – Frye/Daubert—State-by-State

Appendix 2 – Sample Orders for Criminal Discovery

Appendix 3 – Biographies

The Bench Book gives some good advice in very general terms about the need to consider study validity,[1] and to approach scientific evidence with care and “healthy skepticism.”[2] When the Bench Book attempts to instruct on what it represents the scientific method of hypothesis testing, the good advice unravels:

“A scientific hypothesis simply cannot be proved. Statisticians attempt to solve this dilemma by adopting an alternate [sic] hypothesis – the null hypothesis. The null hypothesis is the opposite of the scientific hypothesis. It assumes that the scientific hypothesis is not true. The researcher conducts a statistical analysis of the study data to see if the null hypothesis can be rejected. If the null hypothesis is found to be untrue, the data support the scientific hypothesis as true.”[3]

Even in experimental settings, a statistical analysis of the data do not lead to a conclusion that the null hypothesis is untrue, as opposed to not reasonably compatible with the study’s data. In observational studies, the statistical analysis must acknowledge whether and to what extent the study has excluded bias and confounding. When the Bench Book turns to speak of statistical significance, more trouble ensues:

“The goal of an experiment, or observational study, is to achieve results that are statistically significant; that is, not occurring by chance.”[4]

In the world of result-oriented science, and scientific advocacy, it is perhaps true that scientists seek to achieve statistically significant results. Still, it seems crass to come right out and say so, as opposed to saying that the scientists are querying the data to see whether they are compatible with the null hypothesis. This first pass at statistical significance is only mildly astray compared with the Bench Book’s more serious attempts to define statistical significance and confidence intervals:

“4.10 Statistical Significance

The research field agrees that study outcomes must demonstrate they are not the result of random chance. Leaving room for an error of .05, the study must achieve a 95% level of confidence that the results were the product of the study. This is denoted as p ≤ 05. (or .01 or .1).”[5]

and

“The confidence interval is also a way to gauge the reliability of an estimate. The confidence interval predicts the parameters within which a sample value will fall. It looks at the distance from the mean a value will fall, and is measured by using standard deviations. For example, if all values fall within 2 standard deviations from the mean, about 95% of the values will be within that range.”[6]

Of course, the interval speaks to the precision of the estimate, not its reliability, but that is a small point. These definitions are virtually guaranteed to confuse judges into conflating statistical significance and the coefficient of confidence with the legal burden of proof probability.

The Bench Book runs into problems in interpreting legal decisions, which would seem softer grist for the judicial mill. The authors present dictum from the Daubert decision as though it were a holding:[7]

“As noted in Daubert, ‘[t]he focus, of course, must be solely on principles and methodology, not on the conclusions they generate’.”

The authors fail to mention that this dictum was abandoned in Joiner, and that it is specifically rejected by statute, in the 2000 revision to the Federal Rule of Evidence 702.

Early in the Bench Book, it authors present a subsection entitled “The Myth of Scientific Objectivity,” which they might have borrowed from Feyerabend or Derrida. The heading appears misleading because the text contradicts it:

“Scientists often develop emotional attachments to their work—it can be difficult to abandon an idea. Regardless of bias, the strongest intellectual argument, based on accepted scientific hypotheses, will always prevail, but the road to that conclusion may be fraught with scholarly cul-de-sacs.”[8]

In a similar vein, the authors misleadingly tell readers that “the forefront of science is rarely encountered in court,” and so “much of the science mentioned there shall be considered established….”[9] Of course, the reality is that many causal claims presented in court have already been rejected or held to be indeterminate by the scientific community. And just when readers may think themselves safe from the goblins of nihilism, the authors launch into a theory of naïve probabilism that science is just placing subjective probabilities upon data, based upon preconceived biases and beliefs:

“All of these biases and beliefs play into the process of weighing data, a critical aspect of science. Placing weight on a result is the process of assigning a probability to an outcome. Everything in the universe can be expressed in probabilities.”[10]

So help the expert witness who honestly (and correctly) testifies that the causal claim or its rejection cannot be expressed as a probability statement!

Although I have not read all of the Bench Book closely, there appears to be no meaningful discussion of Rule 703, or of the need to access underlying data to ensure that the proffered scientific opinion under scrutiny has used appropriate methodologies at every step in its development. Even a 412 text cannot address every issue, but this one does little to help the judicial reader find more in-depth help on statistical and scientific methodological issues that arise in occupational and environmental disease claims, and in pharmaceutical products litigation.

The organizations involved in this Bench Book appear to be honest brokers of remedial education for judges. The writing of this Bench Book was funded by the State Justice Institute (SJI) Which is a creation of federal legislation enacted with the laudatory goal of improving the quality of judging in state courts.[11] Despite its provenance in federal legislation, the SJI is a a private, nonprofit corporation, governed by 11 directors appointed by the President, and confirmed by the Senate. A majority of the directors (six) are state court judges, one state court administrator, and four members of the public (no more than two from any one political party). The function of the SJI is to award grants to improve judging in state courts.

The National Judicial College (NJC) originated in the early 1960s, from the efforts of the American Bar Association, American Judicature Society and the Institute of Judicial Administration, to provide education for judges. In 1977, the NJC became a Nevada not-for-profit (501)(c)(3) educational corporation, which its campus at the University of Nevada, Reno, where judges could go for training and recreational activities.

The Justice Speakers Institute appears to be a for-profit company that provides educational resources for judge. A Press Release touts the Bench Book and follow-on webinars. Caveat emptor.

The rationale for this Bench Book is open to question. Unlike the Reference Manual for Scientific Evidence, which was co-produced by the Federal Judicial Center and the National Academies of Science, the Bench Book’s authors are lawyers and judges, without any subject-matter expertise. Unlike the Reference Manual, the Bench Book’s chapters have no scientist or statistician authors, and it shows. Remarkably, the Bench Book does not appear to cite to the Reference Manual or the Manual on Complex Litigation, at any point in its discussion of the federal law of expert witnesses or of scientific or statistical method. Perhaps taxpayers would have been spared substantial expense if state judges were simply encouraged to read the Reference Manual.

[1] Bench Book at 190.

[2] Bench Book at 174 (“Given the large amount of statistical information contained in expert reports, as well as in the daily lives of the general society, the ability to be a competent consumer of scientific reports is challenging. Effective critical review of scientific information requires vigilance, and some healthy skepticism.”).

[3] Bench Book at 137; see also id. at 162.

[4] Bench Book at 148.

[5] Bench Book at 160.

[6] Bench Book at 152.

[7] Bench Book at 233, quoting Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579, 595 (1993).

[8] Bench Book at 10.

[9] Id. at 10.

[10] Id. at 10.

[11] See State Justice Institute Act of 1984 (42 U.S.C. ch. 113, 42 U.S.C. § 10701 et seq.).

Posted in Causation, Evidence, Expert Witness Discovery, Expert Witnesses, Frye, Reference Manual on Scientific Evidence, Rule 702, Rule 703, Scientific Evidence, statistical evidence | Comments Off on Science Bench Book for Judges

Has the American Statistical Association Gone Post-Modern?

March 24th, 2019

Last week, the American Statistical Association (ASA) released a special issue of its journal, The American Statistician, with 43 articles addressing the issue of “statistical significance.” If you are on the ASA’s mailing list, you received an email announcing that

“the lead editorial calls for abandoning the use of ‘statistically significant’, and offers much (not just one thing) to replace it. Written by Ron Wasserstein, Allen Schirm, and Nicole Lazar, the co-editors of the special issue, ‘Moving to a World Beyond ‘p < 0.05’ summarizes the content of the issue’s 43 articles.”

In 2016, the ASA issued its “consensus” statement on statistical significance, in which it articulated six principles for interpreting p-values, and for avoiding erroneous interpretations. Ronald L. Wasserstein & Nicole A. Lazar, “The ASA’s Statement on p-Values: Context, Process, and Purpose,” 70 The American Statistician 129 (2016) [ASA Statement] In the final analysis, that ASA Statement really did not change very much, and could be read fairly only to state that statistical significance was not sufficient for causal inference.¹ Aside from overzealous, over-claiming lawyers and their expert witnesses, few scientists or statisticians had ever maintained that statistical significance was sufficient to support causal inference. Still, many “health effect claims” involve alleged causation that is really a modification of a base rate of a disease or disorder that happens without the allegedly harmful exposure, and which does not invariably happen even with the exposure. It is hard to imagine drawing an inference of such causation without ruling out random error, as well as bias and confounding.

According to the lead editorial for the special issue:

“The ASA Statement on P-Values and Statistical Significance stopped just short of recommending that declarations of ‘statistical significance’ be abandoned. We take that step here. We conclude, based on our review of the articles in this special issue and the broader literature, that it is time to stop using the term ‘statistically significant’ entirely. Nor should variants such as ‘significantly different’, ‘p < 0.05’, and ‘nonsignificant’ survive, whether expressed in words, by asterisks in a table, or in some other way.”²

The ASA (through Wasserstein and colleagues) appear to be condemning dichotomizing p-values, which are a continuum between zero and one. Presumably saying that a p-value is less than 5% is tantamount to dichotomizing, but providing the actual value of the p-value would cause no offense, as long as it was not labeled “significant.”

So although the ASA appears to have gone “whole hog,” the Wasserstein editorial does not appear to condemn assessing random error, or evaluating the extent of random error as part of assessing a study’s support for an association. Reporting p < 0.05 as opposed to p = a real number between zero and one is largely an artifact of statistical tables in the pre-computer era.

So what is the ASA affirmatively recommending? “Much, not just one thing?” Or too much of nothing, which we know makes a man feel ill at ease. Wasserstein’s editorial earnestly admits that there is no replacement for:

“the outsized role that statistical significance has come to play. The statistical community has not yet converged on a simple paradigm for the use of statistical inference in scientific research—and in fact it may never do so.”³

The 42 other articles in the special issue certainly do not converge on any unified, coherent response to the perceived crisis. Indeed, a cursory review of the abstracts alone suggests deep disagreements over an appropriate approach to statistical inference. The ASA may claim to be agnostic in the face of the contradictory recommendations, but there is one thing we know for sure: over-reaching litigants and their expert witnesses will exploit the real or apparent chaos in the ASA’s approach. The lack of coherent, consistent guidance will launch a thousand litigation ships, with no epistemic compass.⁴

1 “The American Statistical Association’s Statement on and of Significance” (March 17, 2016); “The American Statistical Association Statement on Significance Testing Goes to Court – Part I” (Nov. 13, 2018).

2 Ronald L. Wasserstein, Allen L. Schirm, and Nicole A. Lazar, “Editorial: Moving to a World Beyond ‘p < 0.05’,” 73 Am. Statistician S1, S2 (2019).

3 Id. at S2.

4 See, e.g., John P. A. Ioannidis, “Retiring statistical significance would give bias a free pass,” 567 Nature 461 (2019); Valen E. Johnson, “Raise the Bar Rather than Retire Significance,” 567 Nature 461 (2019).

Posted in Rule 702, statistical evidence | Comments Off on Has the American Statistical Association Gone Post-Modern?

Lipitor Diabetes MDL’s Inexact Analysis of Fisher’s Exact Test

March 23rd, 2019

Muriel Bristol was a biologist who studied algae at the Rothamsted Experimental Station in England, after World War I. In addition to her knowledge of plant biology, Bristol claimed the ability to tell whether tea had been added to milk, or the tea poured first and then milk had been added. Bristol, as a scientist and a proper English woman, preferred the latter.

Ronald Fisher, who also worked at Rothamsted, expressed his skepticism over Dr. Bristol’s claim. Fisher set about to design a randomized experiment that would efficiently and effectively test her claim. Bristol was presented with eight cups of tea, four of which were prepared with milk added to tea, and four prepared with tea added to milk. Bristol, of course, was blinded to which was which, but was required to label each according to its manner of preparation. Fisher saw his randomized experiment as a 2 x 2 contingency table, from he could calculate the observed outcome (and ones more extreme if there were any more extreme outcomes) using the assumption of fixed marginal rates and the hypergeometric probability distribution. Fisher’s Exact Test was born at tea time.[1]

Fisher described the origins of his Exact Test in one of his early texts, but he neglected to report whether his experiment vindicated Bristol’s claim. According to David Salsburg, H. Fairfield Smith, one of Fisher’s colleagues, acknowledged that Bristol nailed Fisher’s Exact test, with all eight cups correctly identified. The test has gone on to become an important tool in the statistician’s armamentarium.

Fisher’s Exact, like any statistical test, has model assumptions and preconditions. For one thing, the test is designed for categorical data, with binary outcomes. The test allows us to evaluate whether two proportions are likely different by chance alone, by calculating the probability of the observed outcome, as well as more extreme outcomes.

The calculation of an exact attained significance probability, using Fisher’s approach, provides a one-sided p-value, with no unique solution to calculating a two-side attained significance probability. In discrimination cases, the one-sided p-value may well be more appropriate for the issue at hand. The Fisher’s Exact Test has thus played an important role in showing the judiciary that small sample size need not be an insuperable barrier to meaningful statistical analysis. In discrimination cases, the one-sided p-value provided by the test is not a particular problem.[2]

The difficulty of using Fisher’s Exact for small sample sizes is that the hypergeometric distribution, upon which the test is based, is highly asymmetric. The observed one-sided p-value does not measure the probability of a result equally extreme in the opposite direction. There are at least three ways to calculate the p-value:

Double the one-sided p-value.
Add the point probabilities from the opposite tail that are more extreme than the observed point probability.
Use the mid-P value; that is, add all values more extreme (smaller) than the observed point probability from both sides of the distribution, PLUS ½ of the observed point probability.

Some software programs will proceed in one of these ways by default, but their doing so does guarantee the most accurate measure of two-tailed significance probability.

In the Lipitor MDL for diabetes litigation, Judge Gergel generally used sharp analyses to cut through the rancid fat of litigation claims, to get to the heart of the matter. By and large, he appears to have done a splendid job. In course of gatekeeping under Federal Rule of Evidence 702, however, Judge Gergel may have misunderstood the nature of Fisher’s Exact Test.

Nicholas Jewell is a well-credentialed statistician at the University of California. In the courtroom, Jewell is a well-known expert witness for the litigation industry. He is no novice at generating unreliable opinion testimony. See In re Zoloft Prods. Liab. Litig., No. 12–md–2342, 2015 WL 7776911 (E.D. Pa. Dec. 2, 2015) (excluding Jewell’s opinions as scientifically unwarranted and methodologically flawed). In re Zoloft Prod. Liab. Litig., MDL NO. 2342, 12-MD-2342, 2016 WL 1320799 (E.D. Pa. April 5, 2016) (granting summary judgment after excluding Dr. Jewell). See “The Education of Judge Rufe – The Zoloft MDL” (April 9, 2016).

In the Lipitor cases, some of Jewell’s opinions seemed outlandish indeed, and Judge Gergel generally excluded them. See In re Lipitor Marketing, Sales Practices and Prods. Liab. Litig., 145 F.Supp. 3d 573 (D.S.C. 2015), reconsideration den’d, 2016 WL 827067 (D.S.C. Feb. 29, 2016). As Judge Gergel explained, Jewell calculated a relative risk for abnormal blood glucose in a Lipitor group to be 3.0 (95% C.I., 0.9 to 9.6), using STATA software. Also using STATA, Jewell obtained an attained significance probability of 0.0654, based upon Fisher’s Exact Test. Lipitor Jewell at *7.

Judge Gergel did not report whether Jewell’s reported p-value of 0.0654, was one- or two-sided, but he did state that the attained probability “indicates a lack of statistical significance.” Id. & n. 15. The rest of His Honor’s discussion of the challenged opinion, however, makes clear that of 0.0654 must have been a two-sided value. If it had been a one-sided p-value, then there would have been no way of invoking the mid-p to generate a two-sided p-value below 5%. The mid-p will always be larger than the one-tailed exact p-value generated by Fisher’s Exact Test.

The court noted that Dr. Jewell had testified that he believed that STATA generated this confidence interval by “flip[ping]” the Taylor series approximation. The STATA website notes that it calculates confidence intervals for odds ratios (which are different from the relative risk that Jewell testified he computed), by inverting the Fisher exact test.[3] Id. at *7 & n. 17. Of course, this description suggests that the confidence interval is not based upon exact methods.

STATA does not provide a mid p-value calculation, and so Jewell used an on-line calculator, to obtain a mid p-value of 0.04, which he declared statistically significant. The court took Jewell to task for using the mid p-value as though it were a different analysis or test. Id. at *8. Because the mid-p value will always be larger than the one-sided exact p-value from Fisher’s Exact Test, the court’s explanation does not really make sense:

“Instead, Dr. Jewell turned to the mid-p test, which would ‘[a]lmost surely’ produce a lower p-value than the Fisher exact test.”

Id. at *8. The mid-p test, however, is not different from the Fisher’s exact; rather it is simply a way of dealing with the asymmetrical distribution that underlies the Fisher’s exact, to arrive at a two-tailed p-value that more accurately captures the rate of Type I error.

The MDL court acknowledged that the mid-p approach, was not inherently unreliable, but questioned Jewell’s inconsistent, selective use of the approach for only one test.[4] Jewell certainly did not help the plaintiffs’ cause and his standing by having discarding the analyses that were not incorporated into his report, thus leaving the MDL court to guess at how much selection went on in his process of generating his opinions.. Id. at *9 & n. 19.

None of Jewell’s other calculated p-values involved the mid-p approach, but the court’s criticism begs the question whether the other p-values came from a Fisher’s Exact Test with small sample size, or other highly asymmetrical distribution. Id. at *8. Although Jewell had shown himself willing to engage in other dubious, result-oriented analyses, Jewell’s use of the mid-p for this one comparison may have been within acceptable bounds after all.

The court also noted that Jewell had obtained the “exact p-value and that this p-value was not significant.” Id. The court’s notation here, however, does not report the important detail whether that exact, unreported p-value was merely the doubled of the one-sided p-value given by the Fisher’s Exact Test. As the STATA website, cited by the MDL court, explains:

“The test naturally gives a one-sided p-value, and there are at least four different ways to convert it to a two-sided p-value (Agresti 2002, 93). One way, not implemented in Stata, is to double the one-sided p-value; doubling is simple but can result in p-values larger than one.”

Wesley Eddings, “Fisher’s exact test two-sided idiosyncrasy” (Jan. 2009) (citing Alan Agresti, Categorical Data Analysis 93 (2d ed. 2002)).

On plaintiffs’ motion for reconsideration, the MDL court reaffirmed its findings with respect to Jewell’s use of the mid-p. Lipitor Jewell Reconsidered at *3. In doing so, the court insisted that the one instance in which Jewell used the mid-p stood in stark contrast to all the other instances in which he had used Fisher’s Exact Test. The court then cited to the record to identify 21 other instances in which Jewell used a p-value rather than a mid-p value. The court, however, did not provide the crucial detail whether these 21 other instances actually involved small-sample applications of Fisher’s Exact Test. As result-oriented as Jewell can be, it seems safe to assume that not all his statistical analyses involved Fisher’s Exact Test, with its attendant ambiguity for how to calculate a two-tailed p-value.

[1] Sir Ronald A. Fisher, The Design of Experiments at chapter 2 (1935); see also Stephen Senn, “Tea for three: Of infusions and inferences and milk in first,” Significance 30 (Dec. 2012); David Salsburg, The Lady Tasting Tea: How Statistics Revolutionized Science in the Twentieth Century (2002).

[2] See, e.g., Dendy v. Washington Hosp. Ctr., 431 F. Supp. 873 (D.D.C. 1977) (denying preliminary injunction), rev’d, 581 F.2d 99 (D.C. Cir. 1978) (reversing denial of relief, and remanding for reconsideration). See also National Academies of Science, Reference Manual on Scientific Evidence 255 n.108 (3d ed. 2011) (“Well-known small sample techniques [for testing significance and calculating p-values] include the sign test and Fisher’s exact test.”).

[3] See Wesley Eddings, “Fisher’s exact test two-sided idiosyncrasy” (Jan. 2009), available at <http://www.stata.com/support/faqs/statistics/fishers-exact-test/>, last visited April 19, 2016 (“Stata’s exact confidence interval for the odds ratio inverts Fisher’s exact test.”). This article by Eddings contains a nice discussion of why the Fisher’s Exact Test attained significance probability disagrees with the calculated confidence interval. Eddings points out the asymmetry of the hypergeometric distribution, which complicates arriving at an exact p-value for a two-sided test.

[4] See Barber v. United Airlines, Inc., 17 Fed. Appx. 433, 437 (7th Cir. 2001) (“Because in formulating his opinion Dr. Hynes cherry-picked the facts he considered to render an expert opinion, the district court correctly barred his testimony because such a selective use of facts fails to satisfy the scientific method and Daubert.”).

Posted in Rule 702, statistical evidence | Comments Off on Lipitor Diabetes MDL’s Inexact Analysis of Fisher’s Exact Test