Here is a new commentary from the journal Science Communication 'Forcing a Deterministic Frame on Probabilistic Phenomena: A Communication Blind Spot in Media Coverage of the “Replication Crisis”': doi.org/10.1177/107554…
Read the full article here: t.ly/ylZgM

Here is a shorter version of 'There's Not Much Science in Science', given in a slightly different context, for the International Conference on Using Epidemiological Studies in Health Risk Assessments, Berlin Nov. 2023:
youtube.com/watch?v=fzDyrB…

One of the main issues is that those who most need certain information are less likely to reach or choose to consult the sources of such information. Nevertheless, this seminar is definitely among the knowledge that we all need.

Thanks to Sander Greenland youtube.com/watch?v=N7-yn5…

Check out this new JAMA Network opinion piece from Miguel Hernán Sander Greenland:

“This … proposal is a way to operationalize long-standing calls for shifting research away from testing for effects to estimating effects.”

jamanetwork.com/journals/jama/…

Don't like p-values? We're presenting a completely new interpretation in the context of clinical trials. With Andrew Gelman, Sander Greenland, guido imbens Simon Schwab and Steve Goodman. Sander Greenland Steven Goodman NEJM Evidence
evidence.nejm.org/stoken/default…

So true: 'a collection of perfectly-rational learners can in fact ratchet themselves into believing foolish things...agents with most extreme biases tend to dominate how the system evolves' -
as commonly interpreted, 'statistical significance'/'confidence'/NHST form a case study.

Final print version of my rejoinder to comments on 'Divergence vs. decision P-values' is at
onlinelibrary.wiley.com/doi/10.1111/sj…
All commentators agreed the main distinction is important, but raised key issues, esp. the need to be alert to insensitivities of a diagnostic to model violations.

Nice vignette! I would only change 'the coin is fair' to 'the coin tossing is fair': A skilled tosser can alter the frequency of heads if they know the initial side up and catch the toss. That bias can be avoided by tossing the coin into a wall and letting it land on the floor.

R^2 etc is part of my 'statistics is pants' lecture that I have never given but really should write up so I can let the demons out at last. Just so many aspects of stats are historical baggage applied and reapplied to the wrong contexts and goals. Science before statistics.

For those interested in the defects of 'standardized' coefficients and proportional-variance measures like R-squared and correlation coefficients, see
larspsyll.wordpress.com/2023/05/23/the…

A new article responding to discussants of my SJoS article on divergence vs decision P-values, and adding more general discussion of the limitations of statistics for scientific inference (such as false precision): arxiv.org/abs/2304.01392
initial article: arxiv.org/abs/2301.02478

For those interested in the topic of statistical models vs reality, have a look at Stark, 'Pay No Attention to the Model Behind the Curtain'
link.springer.com/article/10.100…

This commentary by Lars Syll applies with most any topic substituted for economics:
larspsyll.wordpress.com/2022/11/30/the…

Here's an item of English that Neyman got right over Fisher (using 'tested' instead of 'null hypothesis'), maybe because English was Neyman's 4th (?) language & he knew how to write clearly for non-native speakers. Yet the stat community ignored his wisdom amstat.tandfonline.com/doi/suppl/10.1…

'If the sample were made enormous enough, the CI would be almost a spike at the point estimate. Yet even modest systematic errors would offset that spike from the true value and ensure the CI would not contain it. Thus any confidence placed in the CI would be false confidence.'

'I think it is crucially important to underscore that compatibility ('confidence') intervals CI only capture uncertainty traceable to sample-size limitations in the form of 'random error'.'

Courtesy of Sander Greenland Sander Greenland

1/2

For those interested in the long-running issue of confounding vs. noncollapsibility, here's a new thread (joining others on the topic) at Datamethods:
discourse.datamethods.org/t/confounding-…

Our new commentary: Discuss practical importance of results based on interval estimates and p-value functions, not only on point estimates and null p-values.

Old news that have apparently not yet spread very far.

journals.sagepub.com/doi/10.1177/02…

Our new paper: The usual array of dichotomized hypotheses in the introductions of papers makes no sense, at least not in ecology and evolutionary biology.

onlinelibrary.wiley.com/doi/10.1111/je…

For those interested in the ongoing multiple-comparisons controversy, a review of 'When to Adjust Alpha During Multiple Testing' by Mark Rubin has been posted at Qeios, qeios.com/read/D4ZMIZ