Having served as a statistics expert witness, I have some thoughts. The emphasis of this discussion seems to be on the fact that a microscopic p-value is consistent with a difference in populations but doesn't indicate that election cheating is the explanation. Of course, this is true. Also true is that lawyers on both sides are advocates who will choose to use words harmonious with client interest. Attorneys gladly point out that it is not their duty to make the other side's case, but there is a duty to point out when the attorney or expert witness for the other side over-reaches with an explanation. My experience in several age, race and sex discrimination suits, and in cases in which students are suspected of having cheated on multiple choice exams is that there almost always is both statistical and non-statistical evidence to be weighed as a package in order to render a verdict. The one exception I recall is an age discrimination suit in which the p-value associated with the fact that only workers aged 60 or over had been terminated from the workforce was .000000000088 for an age-neutral process. The day after I presented the analysis in a deposition, the plaintiffs' lawyer told me that the case was over – settled out-of-court – because the company didn't want the p-value to be aired in court.
Original Message:
Sent: 12-14-2020 15:20
From: Jonathan Siegel
Subject: Use of Statistical Analysis in Supreme Court Election Case
On Friday, a second "supplemental declaration" from Dr. Cicchetti was filed as an appendix to the State of Texas' reply brief. In his second declaration, Dr. Cicchetti clarified his findings and comments and the inferences he believed should be drawn from them, and responded directly to critics who had called his analysis "nonsense."
in my previous email, I had been careful to distinguish between the findings and conclusions reported by Dr. Cicchetti's himself in his own report, and the statement made by Texas' lawyers aboit his report in their briefs. In doing so, I gave Dr. Cicchetti a certain benefit of the doubt in limiting his responsibility for what the lawyers said about his work.
Dr. Cicchetti's second declaration, unfortunately, removes some of the distinction between his and the lawyers' comments, bringing his position closer to the lawyers. I must therefore regretfully report that, in choosing to speak up and not to remain, Dr. Cicchetti has removed some of the benefit of the doubt I had previously accorded him.
Dr. Cicchetti's second declaration characterizes Biden's win as an "unusual statistical anomaly" and states that "further investigation and audits should be consucted before finalizing the outcome" of the election. Thus, this second declaration makes clear that itiz Dr. Cicchetti himself, and not just the lawyers, who is characterizing his findings as evidence that there is something fishy about the election results warranting Texas' request for the Suprme Court to halt finalization the results pending further investigation and audits.
Dr. Ciccetti also presents additional evidence in support of these conclusions, including evidence of shifts in support in various aubgroups between the 2016 and 2020 elections.
I must therefore report that my concern here lies not just with Dr. Cicchetti's silence regarding the lawyers use of his statements, but his affirmative statements.
I had previously spoken regarding what I had perceived as ethical issues raised bu Dr. Cicchetti's silence regarding the use of his research by others. "Qui tacit consentire," Robert Bolt's Sir Thomas More informs us, "Silence gives consent." While this is so, there is still room for ethical doubt. The ethical person may, in extreme circumstances, sometimes have need to use the cover of silence, and rest in the shelter of the shadow of doubt it casts, as More's case well illustrates. Sometimes silence is the better part of valor.
But here Dr. Cicchetti has spoken. All doubt, sadly, has been removed.
As before, my opinion here is strictly my own.
Here is the link to the second declaration. It begins on p. 31 of the pdf, labeled p. 152a:
https://www.supremecourt.gov/DocketPDF/22/22O155/163497/20201211110907446_TX-v-State-LeaveReply-2020-12-11.pdf
| Supremecourt |
remove preview |
|
|
------------------------------
Jonathan Siegel
Director Clinical Statistics
Original Message:
Sent: 12-09-2020 09:25
From: Jonathan Siegel
Subject: Use of Statistical Analysis in Supreme Court Election Case
The recently filed case of Texas v. Pennsylvania, filed in the Supreme Court's original jurisdiction, included a declaration by Dr. Charles J. Cicchetti, PhD, concerning a statistical analysis of election results. Dr. Cicchetti compared the 2020 election results in the four swing states being sued. (See the appendix in the first link below). He concluded, not surprisingly, I think, that the increase in Democratic votes between the election reflected a real difference in populations, and the probablility of such a difference arising by sampling error from the same population was extremely small.
He similarly compared ballots received on election day with later mail-in ballots and likewise found the probability of their arising by chance from the same population as extremely low: earlier and later ballots came from different populations.
The Supreme Court filing took these results and presented them as evidence that something was wrong with the election. As they said in their main filing (the 2nd link below), "Expert analysis using a commonly accepted statistical test further raises serious questions as to the integrity of this election.)
In their motion for expedition (the first link), they presented Dr. Cicchetti's results as showing that "the probability of former Vice President Biden winning the popular vote in the four defendent states" was astronomically low.
Dr. Cicchetti's declaration doesn't contain a conclusion that the integrity of the election was questionable. It merely shows that in both analyses, the results very likely came from two different populations. The declaration even says that there are many possible reasons why this might be the case. The declaration also presents Dr. Cicchetti as an economist, not a statistician.
But what can or should a statistician, as distinct from some other kind of quantitative professional, ethically do in this situation?
My understanding is that a statistician has an ethical obligation not merely to provide a report of analyses that does not contain misleading conclusions. Rather, if results from that report are published or presented to the public or key decision makers, a statistician has an ethical obligation to ensure that their name is not associated with incorrect or misleading inferences and conclusions.
And I would also think a Supreme Court brief, or indeed any court filing, would count as a publication Thus, I would think a professional statistician would have some obligation to have their name taken off such a brief and not allow their work to support such an inference, and if necessary to write the Supreme Court that their name was used without their permission.
What do you think? What would you have done in this case? What are a statistician's ethical obligations here?
The views in this post are mine only, and do not reflect my employer.
Motion for Expedition with declaration:
https://www.supremecourt.gov/DocketPDF/22/22O155/163048/20201208132827887_TX-v-State-ExpedMot%202020-12-07%20FINAL.pdf
Main Brief (Motion for Leave):
https://www.supremecourt.gov/DocketPDF/22/22O155/162953/20201207234611533_TX-v-State-Motion-2020-12-07%20FINAL.pdf
@@
------------------------------
Jonathan Siegel
Director Clinical Statistics
------------------------------