# Statistical Inference as Severe Testing

## Popper, Falsiﬁcation and Pseudoscience (Notes from my philstat seminar)

My Phil Stat seminar has been meeting for 4 weeks now, and we’re soon to experiment with a small group of outside participants zooming in (write to us, if you are interested in joining us). I’ve been so busy with the seminar that I haven’t blogged. Have you been following? All the materials are on a continually updated syllabus on this blog (SYLLABUS). We’re up to Excursion 2, Tour II.

Last week, we did something unusual: we read from Popper’s Conjectures and Refutations. I wanted to do this because scientists often appeal to distorted and unsophisticated accounts of Popper, especially in discussing falsification, and what demarcates good science from poor science. While I don’t think Popper made good on his most winning slogans, he gives us many seminal launching-oﬀ points for improved accounts of falsiﬁcation, induction, corroboration, and demarcation. Continue reading

## A statistically significant result indicates H’ (μ > μ’) when POW(μ’) is low (not the other way round)–but don’t ignore the standard error

.

1. New monsters. One of the bizarre facts of life in the statistics wars is that a method from one school may be criticized on grounds that it conflicts with a conception that is the reverse of what that school intends. How is that even to be deciphered? That was the difficult task I set for myself in writing Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars (CUP, 2008) [SIST 2018]. I thought I was done, but new monsters keep appearing. In some cases, rather than see how the notion of severity gets us beyond fallacies, misconstruals are taken to criticize severity! So, for example, in the last couple of posts, here and here, I deciphered some of the better known power howlers (discussed in SIST Ex 5 Tour II) I’m linking to all of this tour (in proofs). Continue reading

## Do “underpowered” tests “exaggerate” population effects? (iv)

.

You will often hear that if you reach a just statistically significant result “and the discovery study is underpowered, the observed effects are expected to be inflated” (Ioannidis 2008, p. 64), or “exaggerated” (Gelman and Carlin 2014). This connects to what I’m referring to as the second set of concerns about statistical significance tests, power and magnitude errors. Here, the problem does not revolve around erroneously interpreting power as a posterior probability, as we saw in the fallacy in this post. But there are other points of conflict with the error statistical tester, and much that cries out for clarification — else you will misunderstand the consequences of some of today’s reforms.. Continue reading

## All She Wrote (so far): Error Statistics Philosophy: 10 years on

Dear Reader: I began this blog 10 years ago (Sept. 3, 2011)! A double celebration is taking place at the Elbar Room–remotely for the first time due to Covid– both for the blog and the 3 year anniversary of the physical appearance of my book: Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars [SIST] (CUP, 2018). A special rush edition made an appearance on Sept 3, 2018 in time for the RSS meeting in Cardiff, where we had a session deconstructing the arguments against statistical significance tests (with Sir David Cox, Richard Morey and Aris Spanos). Join us between 7 and 8 pm in a drink of Elba Grease.

.

Many of the discussions in the book were importantly influenced (corrected and improved) by reader’s comments on the blog over the years. I posted several excerpts and mementos from SIST here. I thank readers for their input. Readers might want to look up the topics in SIST on this blog to check out the comments, and see how ideas were developed, corrected and turned into “excursions” in SIST.

I recently invited readers to weigh in on the ASA Task Force on Statistical significance and Replication--any time through September–to be part of a joint guest post (or posts). All contributors will get a free copy of SIST. Continue reading

## Cox’s (1958) Chestnut: You should not get credit (or blame) for something you didn’t do

.

Just as you keep up your physical exercise during the pandemic (sure), you want to keep up with mental gymnastics too. With that goal in mind, and given we’re just a few days from the New Year (and given especially my promised presentation for January 7), here’s one of the two simple examples that will limber you up for the puzzle to ensue. It’s the famous weighing machine example from Sir David Cox (1958)[1]. It is one of the “chestnuts” in the museum exhibits of “chestnuts and howlers” in Excursion 3 (Tour II) of my book Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars (SIST, 2018). So block everything else out for a few minutes and consider 3 pages from SIST …  Continue reading

## Aris Spanos Reviews Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars

A. Spanos

Aris Spanos was asked to review my Statistical Inference as Severe Testing: how to Get Beyond the Statistics Wars (CUP, 2018), but he was to combine it with a review of the re-issue of Ian Hacking’s classic  Logic of Statistical Inference. The journal is OEconomia: History, Methodology, Philosophy. Below are excerpts from his discussion of my book (pp. 843-860). I will jump past the Hacking review, and occasionally excerpt for length.To read his full article go to external journal pdf or stable internal blog pdf. Continue reading

Categories: Spanos, Statistical Inference as Severe Testing

## Cox’s (1958) Chestnut: You shouldn’t get credit (or blame) for something you didn’t do

.

Just as you regularly keep up your physical exercise during the pandemic (sure), you also want to keep up with brain exercise. Given we’re just a few days from New Year’s eve, and given especially that on January 7 I will attempt (for the first time) a highly informal presentation of a controversial result in statistical foundations), here’s a little 2018 marked 60 years since the famous weighing machine example from Sir David Cox (1958)[1]. it is now 61. It’s one of the “chestnuts” in the exhibits of “chestnuts and howlers” in Excursion 3 (Tour II) of my (still) new book Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars (SIST, 2018). It’s especially relevant to take this up now, just before we leave 2019, for reasons that will be revealed over the next day or two. For a sneak preview of those reasons, see the “note to the reader” at the end of this post. So, let’s go back to it, with an excerpt from SIST (pp. 170-173). Continue reading

## The Current State of Play in Statistical Foundations: A View From a Hot-Air Balloon

1.3

Continue to the third, and last stop of Excursion 1 Tour I of Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars (2018, CUP)–Section 1.3. It would be of interest to ponder if (and how) the current state of play in the stat wars has shifted in just one year. I’ll do so in the comments. Use that space to ask me any questions.

How can a discipline, central to science and to critical thinking, have two methodologies, two logics, two approaches that frequently give substantively different answers to the same problems? … Is complacency in the face of contradiction acceptable for a central discipline of science? (Donald Fraser 2011, p. 329)

We [statisticians] are not blameless … we have not made a concerted professional eﬀort to provide the scientific world with a uniﬁed testing methodology. (J. Berger 2003, p. 4)

Categories: Statistical Inference as Severe Testing

## Severity: Strong vs Weak (Excursion 1 continues)

1.2

Marking one year since the appearance of my book: Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars (2018, CUP), let’s continue to the second stop (1.2) of Excursion 1 Tour 1. It begins on p. 13 with a quote from statistician George Barnard. Assorted reflections will be given in the comments. Ask me any questions pertaining to the Tour.

• I shall be concerned with the foundations of the subject. But in case it should be thought that this means I am not here strongly concerned with practical applications, let me say right away that confusion about the foundations of the subject is responsible, in my opinion, for much of the misuse of the statistics that one meets in ﬁelds of application such as medicine, psychology, sociology, economics, and so forth. (George Barnard 1985, p. 2)
Categories: Statistical Inference as Severe Testing

## How My Book Begins: Beyond Probabilism and Performance: Severity Requirement

This week marks one year since the general availability of my book: Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars (2018, CUP). Here’s how it begins (Excursion 1 Tour 1 (1.1)). Material from the preface is here. I will sporadically give some “one year later” reflections in the comments. I invite readers to ask me any questions pertaining to the Tour.

The journey begins..(1.1)

I’m talking about a speciﬁc, extra type of integrity that is [beyond] not lying, but bending over backwards to show how you’re maybe wrong, that you ought to have when acting as a scientist. (Feynman 1974/1985, p. 387)

It is easy to lie with statistics. Or so the cliché goes. It is also very diﬃcult to uncover these lies without statistical methods – at least of the right kind. Self- correcting statistical methods are needed, and, with minimal technical fanfare, that’s what I aim to illuminate. Since Darrell Huﬀ wrote How to Lie with Statistics in 1954, ways of lying with statistics are so well worn as to have emerged in reverberating slogans:

• Association is not causation.
• Statistical signiﬁcance is not substantive signiﬁcamce
• No evidence of risk is not evidence of no risk.
• If you torture the data enough, they will confess.

## SIST: All Excerpts and Mementos: May 2018-June 2020 (updated)

Introduction & Overview

The Meaning of My Title: Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars* 05/19/18

Blurbs of 16 Tours: Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars (SIST) 03/05/19

Statistical Inference as Severe Testing: Preface

Excursion 1

EXCERPTS

Tour I Ex1 TI (full proofs)

Excursion 1 Tour I: Beyond Probabilism and Performance: Severity Requirement (1.1) 09/08/18

Excursion 1 Tour I (2nd stop): Probabilism, Performance, and Probativeness (1.2) 09/11/18

Excursion 1 Tour I (3rd stop): The Current State of Play in Statistical Foundations: A View From a Hot-Air Balloon (1.3) 09/15/18

Tour II Ex1 TII (full proofs)

Excursion 1 Tour II: Error Probing Tools versus Logics of Evidence-Excerpt 04/04/19

Souvenir C: A Severe Tester’s Translation Guide (Excursion 1 Tour II) 11/08/18

MEMENTOS

Tour Guide Mementos (Excursion 1 Tour II of How to Get Beyond the Statistics Wars) 10/29/18

Excursion 2

EXCERPTS

Tour I (full proofs)

Excursion 2: Taboos of Induction and Falsification: Tour I (first stop) 09/29/18

“It should never be true, though it is still often said, that the conclusions are no more accurate than the data on which they are based” (Keepsake by Fisher, 2.1) 10/05/18

Tour II (full proofs)

Excursion 2 Tour II (3rd stop): Falsiﬁcation, Pseudoscience, Induction (2.3) 10/10/18
Ex 2 TII (Full proofs)

MEMENTOS

Tour Guide Mementos and Quiz 2.1 (Excursion 2 Tour I Induction and Confirmation) 11/14/18

Mementos for Excursion 2 Tour II Falsification, Pseudoscience, Induction 11/17/18

Excursion 3

EXCERPTS

Tour I Ex3 TI (full proofs)

Where are Fisher, Neyman, Pearson in 1919? Opening of Excursion 3 11/30/18

Neyman-Pearson Tests: An Episode in Anglo-Polish Collaboration: Excerpt from Excursion 3 (3.2) 12/01/18

First Look at N-P Methods as Severe Tests: Water plant accident [Exhibit (i) from Excursion 3] 12/04/18

Tour II Ex3 TII (full proofs)

It’s the Methods, Stupid: Excerpt from Excursion 3 Tour II (Mayo 2018, CUP) 12/11/18

60 Years of Cox’s (1958) Chestnut: Excerpt from Excursion 3 tour II. 12/29/18

Tour III Ex3 Tour III(full proofs)

Capability and Severity: Deeper Concepts: Excerpts From Excursion 3 Tour III 12/20/18

MEMENTOS

Memento & Quiz (on SEV): Excursion 3, Tour I 12/08/18

Mementos for “It’s the Methods, Stupid!” Excursion 3 Tour II (3.4-3.6) 12/13/18

Tour Guide Mementos From Excursion 3 Tour III: Capability and Severity: Deeper Concepts 12/26/18

Excursion 4

EXCERPTS

Tour I (Full Excursion 4 Tour I)

Excerpt from Excursion 4 Tour I: The Myth of “The Myth of Objectivity” (Mayo 2018, CUP) 12/26/18

Tour II (Full Excursion 4 Tour II)

Excerpt from Excursion 4 Tour II: 4.4 “Do P-Values Exaggerate the Evidence?” 01/10/19

Tour III (Full proofs of Excursion 4 Tour III)

Tour IV (Full proofs of Excursion 4 Tour IV)

Excerpt from Excursion 4 Tour IV: More Auditing: Objectivity and Model Checking 01/27/19

MEMENTOS

Mementos from Excursion 4: Blurbs of Tours I-IV 01/13/19

Excursion 5

Tour I (full proofs)

(Full) Excerpt: Excursion 5 Tour I — Power: Pre-data and Post-data (from “SIST: How to Get Beyond the Stat Wars”) 04/27/19

Tour II (full proofs)

(Full) Excerpt. Excursion 5 Tour II: How Not to Corrupt Power (Power Taboos, Retro Power, and Shpower) 06/07/19

Tour III (full proofs)

Deconstructing the Fisher-Neyman conflict wearing Fiducial glasses + Excerpt 5.8 from SIST 02/23/19

Excursion 6

Tour I (full proofsWhat Ever Happened to Bayesian Foundations?

Tour II (full proofs)

Excerpts: Souvenir Z: Understanding Tribal Warfare +  6.7 Farewell Keepsake from SIST + List of Souvenirs 05/04/19 (full excerpt)

Categories: SIST, Statistical Inference as Severe Testing

## (Full) Excerpt. Excursion 5 Tour II: How Not to Corrupt Power (Power Taboos, Retro Power, and Shpower)

returned from London…

The concept of a test’s power is still being corrupted in the myriad ways discussed in 5.5, 5.6.  I’m excerpting all of Tour II of Excursion 5, as I did with Tour I (of Statistical Inference as Severe Testing:How to Get Beyond the Statistics Wars 2018, CUP)*. Originally the two Tours comprised just one, but in finalizing corrections, I decided the two together was too long of a slog, and I split it up. Because it was done at the last minute, some of the terms in Tour II rely on their introductions in Tour I.  Here’s how it starts:

5.5 Power Taboos, Retrospective Power, and Shpower

Let’s visit some of the more populous tribes who take issue with power – by which we mean ordinary power – at least its post-data uses. Power Peninsula is often avoided due to various “keep out” warnings and prohibitions, or researchers come during planning, never to return. Why do some people consider it a waste of time, if not totally taboo, to compute power once we know the data? A degree of blame must go to N-P, who emphasized the planning role of power, and only occasionally mentioned its use in determining what gets “confirmed” post-data. After all, it’s good to plan how large a boat we need for a philosophical excursion to the Lands of Overlapping Statistical Tribes, but once we’ve made it, it doesn’t matter that the boat was rather small. Or so the critic of post-data power avers. A crucial disanalogy is that with statistics, we don’t know that we’ve “made it there,” when we arrive at a statistically significant result. The statistical significance alarm goes off, but you are not able to see the underlying discrepancy that generated the alarm you hear. The problem is to make the leap from the perceived alarm to an aspect of a process, deep below the visible ocean, responsible for its having been triggered. Then it is of considerable relevance to exploit information on the capability of your test procedure to result in alarms going off (perhaps with different decibels of loudness), due to varying values of the parameter of interest. There are also objections to power analysis with insignificant results. Continue reading

## SIST: All Excerpts and Mementos: May 2018-May 2019

view from a hot-air balloon

Introduction & Overview

The Meaning of My Title: Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars* 05/19/18

Blurbs of 16 Tours: Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars (SIST) 03/05/19

Excursion 1

EXCERPTS

Tour I (full proofs)

Excursion 1 Tour I: Beyond Probabilism and Performance: Severity Requirement (1.1) 09/08/18

Excursion 1 Tour I (2nd stop): Probabilism, Performance, and Probativeness (1.2) 09/11/18

Excursion 1 Tour I (3rd stop): The Current State of Play in Statistical Foundations: A View From a Hot-Air Balloon (1.3) 09/15/18

Tour II (full proofs)

Excursion 1 Tour II: Error Probing Tools versus Logics of Evidence-Excerpt 04/04/19

Souvenir C: A Severe Tester’s Translation Guide (Excursion 1 Tour II) 11/08/18

MEMENTOS

Tour Guide Mementos (Excursion 1 Tour II of How to Get Beyond the Statistics Wars) 10/29/18

Excursion 2

EXCERPTS

Tour I (full proofs)

Excursion 2: Taboos of Induction and Falsification: Tour I (first stop) 09/29/18

“It should never be true, though it is still often said, that the conclusions are no more accurate than the data on which they are based” (Keepsake by Fisher, 2.1) 10/05/18

Tour II (full proofs)

Excursion 2 Tour II (3rd stop): Falsiﬁcation, Pseudoscience, Induction (2.3) 10/10/18

MEMENTOS

Tour Guide Mementos and Quiz 2.1 (Excursion 2 Tour I Induction and Confirmation) 11/14/18

Mementos for Excursion 2 Tour II Falsification, Pseudoscience, Induction 11/17/18

Excursion 3

EXCERPTS

Tour I (full proofs)

Where are Fisher, Neyman, Pearson in 1919? Opening of Excursion 3 11/30/18

Neyman-Pearson Tests: An Episode in Anglo-Polish Collaboration: Excerpt from Excursion 3 (3.2) 12/01/18

First Look at N-P Methods as Severe Tests: Water plant accident [Exhibit (i) from Excursion 3] 12/04/18

Tour II (full proofs)

It’s the Methods, Stupid: Excerpt from Excursion 3 Tour II (Mayo 2018, CUP) 12/11/18

60 Years of Cox’s (1958) Chestnut: Excerpt from Excursion 3 tour II. 12/29/18

Tour III (full proofs)

Capability and Severity: Deeper Concepts: Excerpts From Excursion 3 Tour III 12/20/18

MEMENTOS

Memento & Quiz (on SEV): Excursion 3, Tour I 12/08/18

Mementos for “It’s the Methods, Stupid!” Excursion 3 Tour II (3.4-3.6) 12/13/18

Tour Guide Mementos From Excursion 3 Tour III: Capability and Severity: Deeper Concepts 12/26/18

Excursion 4

EXCERPTS

Tour I (full proofs)

Excerpt from Excursion 4 Tour I: The Myth of “The Myth of Objectivity” (Mayo 2018, CUP) 12/26/18

Tour II (full proofs)

Excerpt from Excursion 4 Tour II: 4.4 “Do P-Values Exaggerate the Evidence?” 01/10/19

Tour III (full proofs)

Tour IV (full proofs)

Excerpt from Excursion 4 Tour IV: More Auditing: Objectivity and Model Checking 01/27/19

MEMENTOS

Mementos from Excursion 4: Blurbs of Tours I-IV 01/13/19

Excursion 5

Tour I (full proofs)

(full) Excerpt: Excursion 5 Tour I — Power: Pre-data and Post-data (from “SIST: How to Get Beyond the Stat Wars”) 04/27/19

Tour II (full proofs)

Tour III (full proofs)

Deconstructing the Fisher-Neyman conflict wearing Fiducial glasses + Excerpt 5.8 from SIST 02/23/19

Excursion 6

Tour I (full proofs)

Tour II (full proofs)

Excerpts: Souvenir Z: Understanding Tribal Warfare +  6.7 Farewell Keepsake from SIST + List of Souvenirs 05/04/19

Categories: SIST, Statistical Inference as Severe Testing

## Excerpts: Final Souvenir Z, Farewell Keepsake & List of Souvenirs

.

We’ve reached our last Tour (of SIST)*: Pragmatic and Error Statistical Bayesians (Excursion 6), marking the end of our reading with Souvenir Z, the final Souvenir, as well as the Farewell Keepsake in 6.7. Our cruise ship Statinfasst, currently here at Thebes, will be back at dock for maintenance for our next launch at the Summer Seminar in Phil Stat (July 28-Aug 11). Although it’s not my preference that new readers begin with the Farewell Keepsake (it contains a few spoilers), I’m excerpting it together with Souvenir Z (and a list of all souvenirs A – Z) here, and invite all interested readers to peer in. There’s a check list on p. 437: If you’re in the market for a new statistical account, you’ll want to test if it satisfies the items on the list. Have fun!

Souvenir Z: Understanding Tribal Warfare

We began this tour asking: Is there an overarching philosophy that “matches contemporary attitudes”? More important is changing attitudes. Not to encourage a switch of tribes, or even a tribal truce, but something more modest and actually achievable: to understand and get beyond the tribal warfare. To understand them, at minimum, requires grasping how the goals of probabilism differ from those of probativeness. This leads to a way of changing contemporary attitudes that is bolder and more challenging. Snapshots from the error statistical lens let you see how frequentist methods supply tools for controlling and assessing how well or poorly warranted claims are. All of the links, from data generation to modeling, to statistical inference and from there to substantive research claims, fall into place within this statistical philosophy. If this is close to being a useful way to interpret a cluster of methods, then the change in contemporary attitudes is radical: it has never been explicitly unveiled. Our journey was restricted to simple examples because those are the ones fought over in decades of statistical battles. Much more work is needed. Those grappling with applied problems are best suited to develop these ideas, and see where they may lead. I never promised,when you bought your ticket for this passage, to go beyond showing that viewing statistics as severe testing will let you get beyond the statistics wars.

6.7 Farewell Keepsake

Despite the eclecticism of statistical practice, conflicting views about the roles of probability and the nature of statistical inference – holdovers from long-standing frequentist–Bayesian battles – still simmer below the surface of today’s debates. Reluctance to reopen wounds from old battles has allowed them to fester. To assume all we need is an agreement on numbers – even if they’re measuring different things – leads to statistical schizophrenia. Rival conceptions of the nature of statistical inference show up unannounced in the problems of scientific integrity, irreproducibility, and questionable research practices, and in proposed methodological reforms. If you don’t understand the assumptions behind proposed reforms, their ramifications for statistical practice remain hidden from you.

Rival standards reflect a tension between using probability (a) to constrain the probability that a method avoids erroneously interpreting data in a series of applications (performance), and (b) to assign degrees of support, confirmation, or plausibility to hypotheses (probabilism). We set sail on our journey with an informal tool for telling what’s true about statistical inference: If little if anything has been done to rule out flaws in taking data as evidence for a claim, then that claim has not passed a severe test . From this minimal severe-testing requirement, we develop a statistical philosophy that goes beyond probabilism and performance. The goals of the severe tester (probativism) arise in contexts sufficiently different from those of probabilism that you are free to hold both, for distinct aims (Section 1.2). For statistical inference in science, it is severity we seek. A claim passes with severity only to the extent that it is subjected to, and passes, a test that it probably would have failed, if false. Viewing statistical inference as severe testing alters long-held conceptions of what’s required for an adequate account of statistical inference in science. In this view, a normative statistical epistemology –  an account of what’ s warranted to infer –  must be:

directly altered by biasing selection effects
able to falsify claims statistically
able to test statistical model assumptions
able to block inferences that violate minimal severity

These overlapping and interrelated requirements are disinterred over the course of our travels. This final keepsake collects a cluster of familiar criticisms of error statistical methods. They are not intended to replace the detailed arguments, pro and con, within; here we cut to the chase, generally keeping to the language of critics. Given our conception of evidence, we retain testing language even when the statistical inference is an estimation, prediction, or proposed answer to a question. The concept of severe testing is sufficiently general to apply to any of the methods now in use. It follows that a variety of statistical methods can serve to advance the severity goal, and that they can, in principle, find their foundations in an error statistical philosophy. However, each requires supplements and reformulations to be relevant to real-world learning. Good science does not turn on adopting any formal tool, and yet the statistics wars often focus on whether to use one type of test (or estimation, or model selection) or another. Meta-researchers charged with instigating reforms do not agree, but the foundational basis for the disagreement is left unattended. It is no wonder some see the statistics wars as proxy wars between competing tribe leaders, each keen to advance one or another tool, rather than about how to do better science. Leading minds are drawn into inconsequential battles, e.g., whether to use a prespecified cut-off  of 0.025 or 0.0025 –  when in fact good inference is not about cut-offs altogether but about a series of small-scale steps in collecting, modeling and analyzing data that work together to find things out. Still, we need to get beyond the statistics wars in their present form. By viewing a contentious battle in terms of a difference in goals –  finding highly probable versus highly well probed hypotheses – readers can see why leaders of rival tribes often talk past each other. To be clear, the standpoints underlying the following criticisms are open to debate; we’re far from claiming to do away with them. What should be done away with is rehearsing the same criticisms ad nauseum. Only then can we hear the voices of those calling for an honest standpoint about responsible science.

1. NHST Licenses Abuses. First, there’s the cluster of criticisms directed at an abusive NHST animal: NHSTs infer from a single P-value below an arbitrary cut-off to evidence for a research claim, and they encourage P-hacking, fishing, and other selection effects. The reply: this ignores crucial requirements set by Fisher and other founders: isolated significant results are poor evidence of a genuine effect and statistical significance doesn’t warrant substantive, (e.g., causal) inferences. Moreover, selective reporting invalidates error probabilities. Some argue significance tests are un-Popperian because the higher the sample size, the easier to infer one’s research hypothesis. It’s true that with a sufficiently high sample size any discrepancy from a null hypothesis has a high probability of being detected, but statistical significance does not license inferring a research claim H. Unless H’s errors have been well probed by merely finding a small P-value, H passes an extremely insevere test. No mountains out of molehills (Sections 4.3 and 5.1). Enlightened users of statistical tests have rejected the cookbook, dichotomous NHST, long lampooned: such criticisms are behind the times. When well-intentioned aims of replication research are linked to these retreads, it only hurts the cause. One doesn’t need a sharp dichotomy to identify rather lousy tests – a main goal for a severe tester. Granted, policy-making contexts may require cut-offs, as do behavioristic setups. But in those contexts, a test’s error probabilities measure overall error control, and are not generally used to assess well-testedness. Even there, users need not fall into the NHST traps (Section 2.5). While attention to banning terms is the least productive aspect of the statistics wars, since NHST is not used by Fisher or N-P, let’s give the caricature its due and drop the NHST acronym; “statistical tests” or “error statistical tests” will do. Simple significance tests are a small part of a conglomeration of error statistical methods.

To continue reading: Excerpt Souvenir Z, Farewell Keepsake & List of Souvenirs can be found here.

*We are reading Statistical Inference as Severe Testing: How to Get beyond the Statistics Wars (2018, CUP)

***

Where YOU are in the journey.

Categories: SIST, Statistical Inference as Severe Testing

## (full) Excerpt: Excursion 5 Tour I — Power: Pre-data and Post-data (from “SIST: How to Get Beyond the Stat Wars”)

S.S. StatInfasST

It’s a balmy day today on Ship StatInfasST: An invigorating wind has a salutary effect on our journey. So, for the first time I’m excerpting all of Excursion 5 Tour I (proofs) of Statistical Inference as Severe Testing How to Get Beyond the Statistics Wars (2018, CUP)

A salutary effect of power analysis is that it draws one forcibly to consider the magnitude of effects. In psychology, and especially in soft psychology, under the sway of the Fisherian scheme, there has been little consciousness of how big things are. (Cohen 1990, p. 1309)

So how would you use power to consider the magnitude of effects were you drawn forcibly to do so? In with your breakfast is an exercise to get us started on today’ s shore excursion.

Suppose you are reading about a statistically signifi cant result x (just at level α ) from a one-sided test T+ of the mean of a Normal distribution with IID samples, and known σ: H0 : μ ≤ 0 against H1 : μ > 0. Underline the correct word, from the perspective of the (error statistical) philosophy, within which power is defined.

• If the test’ s power to detect μ′ is very low (i.e., POW(μ′ ) is low), then the statistically significant x is poor/good evidence that μ > μ′ .
• Were POW(μ′ ) reasonably high, the inference to μ > μ′ is reasonably/poorly warranted.

## Blurbs of 16 Tours: Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars (SIST)

### Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars (2018, CUP)

Deborah G. Mayo

Abstract for Book

By disinterring the underlying statistical philosophies this book sets the stage for understanding and finally getting beyond today’s most pressing controversies revolving around statistical methods and irreproducible findings. Statistical Inference as Severe Testing takes the reader on a journey that provides a non-technical “how to” guide for zeroing in on the most influential arguments surrounding commonly used–and abused– statistical methods. The book sets sail with a tool for telling what’s true about statistical controversies: If little if anything has been done to rule out flaws in taking data as evidence for a claim, then that claim has not passed a stringent or severe test. In the severe testing account, probability arises in inference, not to measure degrees of plausibility or belief in hypotheses, but to assess and control how severely tested claims are. Viewing statistical inference as severe testing supplies novel solutions to problems of induction, falsification and demarcating science from pseudoscience, and serves as the linchpin for understanding and getting beyond the statistics wars. The book links philosophical questions about the roles of probability in inference to the concerns of practitioners in psychology, medicine, biology, economics, physics and across the landscape of the natural and social sciences.

Keywords for book:

Severe testing, Bayesian and frequentist debates, Philosophy of statistics, Significance testing controversy, statistics wars, replication crisis, statistical inference, error statistics, Philosophy and history of Neyman, Pearson and Fisherian statistics, Popperian falsification

Categories: Statistical Inference as Severe Testing

## 4.8 All Models Are False

. . . it does not seem helpful just to say that all models are wrong. The very word model implies simplification and idealization. . . . The construction of idealized representations that capture important stable aspects of such systems is, however, a vital part of general scientific analysis. (Cox 1995, p. 456)

A popular slogan in statistics and elsewhere is “all models are false!”  Is this true? What can it mean to attribute a truth value to a model? Clearly what is meant involves some assertion or hypothesis about the model – that it correctly or incorrectly represents some phenomenon in some respect or to some degree. Such assertions clearly can be true. As Cox observes, “the very word model implies simplification and idealization.”  To declare, “all models are false”  by dint of their being idealizations or approximations, is to stick us with one of those  “all flesh is grass”  trivializations (Section 4.1). So understood, it follows that all statistical models are false, but we have learned nothing about how statistical models may be used to infer true claims about problems of interest. Since the severe tester’s goal in using approximate statistical models is largely to learn where they break down, their strict falsity is a given. Yet it does make her wonder why anyone would want to place a probability assignment on their truth, unless it was 0? Today’s tour continues our journey into solving the problem of induction (Section 2.7). Continue reading

Categories: Statistical Inference as Severe Testing

## Excursion 4: Objectivity and Auditing (blurbs of Tours I – IV)

.

Excursion 4 Tour I: The Myth of “The Myth of Objectivity”

Blanket slogans such as “all methods are equally objective and subjective” trivialize into oblivion the problem of objectivity. Such cavalier attitudes are at odds with the moves to take back science The goal of this tour is to identify what there is in objectivity that we won’t give up, and shouldn’t. While knowledge gaps leave room for biases and wishful thinking, we regularly come up against data that thwart our expectations and disagree with predictions we try to foist upon the world. This pushback supplies objective constraints on which our critical capacity is built. Supposing an objective method is to supply formal, mechanical, rules to process data is a holdover of a discredited logical positivist philosophy.Discretion in data generation and modeling does not warrant concluding: statistical inference is a matter of subjective belief. It is one thing to talk of our models as objects of belief and quite another to maintain that our task is to model beliefs. For a severe tester, a statistical method’s objectivity requires the ability to audit an inference: check assumptions, pinpoint blame for anomalies, falsify, and directly register how biasing selection effects–hunting, multiple testing and cherry-picking–alter its error probing capacities.

Keywords

objective vs. subjective, objectivity requirements, auditing, dirty hands argument, phenomena vs. epiphenomena, logical positivism, verificationism, loss and cost functions, default Bayesians, equipoise assignments, (Bayesian) wash-out theorems, degenerating program, transparency, epistemology: internal/external distinction

Excursion 4 Tour II: Rejection Fallacies: Whose Exaggerating What?

We begin with the Mountains out of Molehills Fallacy (large n problem): The fallacy of taking a (P-level) rejection of H0 with larger sample size as indicating greater discrepancy from H0 than with a smaller sample size. (4.3). The Jeffreys-Lindley paradox shows with large enough n, a .05 significant result can correspond to assigning H0 a high probability .95. There are family feuds as to whether this is a problem for Bayesians or frequentists! The severe tester takes account of sample size in interpreting the discrepancy indicated. A modification of confidence intervals (CIs) is required.

It is commonly charged that significance levels overstate the evidence against the null hypothesis (4.4, 4.5). What’s meant? One answer considered here, is that the P-value can be smaller than a posterior probability to the null hypothesis, based on a lump prior (often .5) to a point null hypothesis. There are battles between and within tribes of Bayesians and frequentists. Some argue for lowering the P-value to bring it into line with a particular posterior. Others argue the supposed exaggeration results from an unwarranted lump prior to a wrongly formulated null.We consider how to evaluate reforms based on bayes factor standards (4.5). Rather than dismiss criticisms of error statistical methods that assume a standard from a rival account, we give them a generous reading. Only once the minimal principle for severity is violated do we reject them. Souvenir R summarizes the severe tester’s interpretation of a rejection in a statistical significance test. At least 2 benchmarks are needed: reports of discrepancies (from a test hypothesis) that are, and those that are not, well indicated by the observed difference.

Keywords

significance test controversy, mountains out of molehills fallacy, large n problem, confidence intervals, P-values exaggerate evidence, Jeffreys-Lindley paradox, Bayes/Fisher disagreement, uninformative (diffuse) priors, Bayes factors, spiked priors, spike and slab, equivocating terms, severity interpretation of rejection (SIR)

Excursion 4 Tour III: Auditing: Biasing Selection Effects & Randomization

Tour III takes up Peirce’s “two rules of inductive inference”: predesignation (4.6) and randomization (4.7). The Tour opens on a court case transpiring: the CEO of a drug company is being charged with giving shareholders an overly rosy report based on post-data dredging for nominally significant benefits. Auditing a result includes checking for (i) selection effects, (ii) violations of model assumptions, and (iii) obstacles to moving from statistical to substantive claims. We hear it’s too easy to obtain small P-values, yet replication attempts find it difficult to get small P-values with preregistered results. I call this the paradox of replication. The problem isn’t P-values but failing to adjust them for cherry picking and other biasing selection effects. Adjustments by Bonferroni and false discovery rates are considered. There is a tension between popular calls for preregistering data analysis, and accounts that downplay error probabilities. Worse, in the interest of promoting a methodology that rejects error probabilities, researchers who most deserve lambasting are thrown a handy line of defense. However, data dependent searching need not be pejorative. In some cases, it can improve severity. (4.6)

Big Data cannot ignore experimental design principles. Unless we take account of the sampling distribution, it becomes difficult to justify resampling and randomization. We consider RCTs in development economics (RCT4D) and genomics. Failing to randomize microarrays is thought to have resulted in a decade lost in genomics. Granted the rejection of error probabilities is often tied to presupposing their relevance is limited to long-run behavioristic goals, which we reject. They are essential for an epistemic goal: controlling and assessing how well or poorly tested claims are. (4.7)

Keywords

error probabilities and severity, predesignation, biasing selection effects, paradox of replication, capitalizing on chance, bayes factors, batch effects, preregistration, randomization: Bayes-frequentist rationale, bonferroni adjustment, false discovery rates, RCT4D, genome-wide association studies (GWAS)

Excursion 4 Tour IV: More Auditing: Objectivity and Model Checking

While all models are false, it’s also the case that no useful models are true. Were a model so complex as to represent data realistically, it wouldn’t be useful for finding things out. A statistical model is useful by being adequate for a problem, meaning it enables controlling and assessing if purported solutions are well or poorly probed and to what degree. We give a way to define severity in terms of solving a problem.(4.8) When it comes to testing model assumptions, many Bayesians agree with George Box (1983) that “it requires frequentist theory of significance tests” (p. 57). Tests of model assumptions, also called misspecification (M-S) tests, are thus a promising area for Bayes-frequentist collaboration. (4.9) When the model is in doubt, the likelihood principle is inapplicable or violated. We illustrate a non-parametric bootstrap resampling. It works without relying on a theoretical  probability distribution, but it still has assumptions. (4.10). We turn to the M-S testing approach of econometrician Aris Spanos.(4.11) I present the high points for unearthing spurious correlations, and assumptions of linear regression, employing 7 figures. M-S tests differ importantly from model selection–the latter uses a criterion for choosing among models, but does not test their statistical assumptions. They test fit rather than whether a model has captured the systematic information in the data.

Keywords

adequacy for a problem, severity (in terms of problem solving), model testing/misspecification (M-S) tests, likelihood principle conflicts, bootstrap, resampling, Bayesian p-value, central limit theorem, nonsense regression, significance tests in model checking, probabilistic reduction, respecification

Where you are in the Journey

Categories: SIST, Statistical Inference as Severe Testing

## Excerpt from Excursion 4 Tour II: 4.4 “Do P-Values Exaggerate the Evidence?”

getting beyond…

### Excerpt from Excursion 4 Tour II*

4.4 Do P-Values Exaggerate the Evidence?

“Significance levels overstate the evidence against the null hypothesis,” is a line you may often hear. Your first question is:

What do you mean by overstating the evidence against a hypothesis?

Several (honest) answers are possible. Here is one possibility:

What I mean is that when I put a lump of prior weight π0 of 1/2 on a point null H0 (or a very small interval around it), the P-value is smaller than my Bayesian posterior probability on H0.

More generally, the “P-values exaggerate” criticism typically boils down to showing that if inference is appraised via one of the probabilisms – Bayesian posteriors, Bayes factors, or likelihood ratios – the evidence against the null (or against the null and in favor of some alternative) isn’t as big as 1 − P. Continue reading

Categories: SIST, Statistical Inference as Severe Testing

## January Invites: Ask me questions (about SIST), Write Discussion Analyses (U-Phils)

.

ASK ME. Some readers say they’re not sure where to ask a question of comprehension on Statistical Inference as Severe Testing: How to Get Beyond the Statistics Wars (2018, CUP)–SIST– so here’s a special post to park your questions of comprehension (to be placed in the comments) on a little over the first half of the book. That goes up to and includes Excursion 4 Tour I on “The Myth of ‘The Myth of Objectivity'”. However,I will soon post on Tour II: Rejection Fallacies: Who’s Exaggerating What? So feel free to ask questions of comprehension as far as p.259.

All of the SIST BlogPost (Excerpts and Mementos) so far are here.

.

WRITE A DISCUSSION NOTE: Beginning January 16, anyone who wishes to write a discussion note (on some aspect or issue up to p. 259 are invited to do so (<750 words, longer if you wish). Send them to my error email.  I will post as many as possible on this blog.

We initially called such notes “U-Phils” as in “You do a Philosophical analysis”, which really only means it’s an analytic excercize that strives to first give the most generous interpretation to positions, and then examines them. See the general definition of  a U-Phil.

Some Examples:

Mayo, Senn, and Wasserman on Gelman’s RMM** Contribution

U-Phil: A Further Comment on Gelman by Christian Hennig.

For a whole group of reader contributions, including Jim Berger on Jim Berger, see: Earlier U-Phils and Deconstructions

If you’re writing a note on objectivity, you might wish to compare and contrast Excursion 4 Tour I with a paper by Gelman and Hennig (2017): “Beyond subjective and objective in Statistics”.

These invites extend through January.

Categories: SIST, Statistical Inference as Severe Testing