As a fan of Edward Tufte and his classic, The Visual Display of Quantitative Information (1983), since grad school, I wanted to pass along the following statistical chart gems (drawn from here).
Wednesday, October 28, 2026
Sunday, October 18, 2026
More Settlement Data: California's STAND Act
In the wake of the "#MeToo" movement, in 2018 California lawmakers enacted the Stand Together Against Non-Disclosure (STAND) Act, which sets out to promote greater transparency in the resolution of sex discrimination, harassment, and abuse cases. At the same time, the STAND Act also helps frame analyses of questions relating to "what exactly happens in the wake of [such a] reform."
In another installment from their on-going exploration of "secret settlements," David Freeman Engstrom (Stanford) et al., assess a quarter-million case filings from the Los Angeles County Superior Court for insights on how the STAND Act may have influenced litigation patterns.
Core claims reported in Shedding Light on Secret Settlements: An Empirical Study of California’s STAND Act include what the study did not find. Notably, "the STAND Act did not yield a sharp increase or decrease in case filings. Nor did the Act appear to significantly prolong cases or amplify their intensity." That is to say, the paper argues that evidence from the STAND Act implies that parties will still settle cases even without the inducement of secrecy. An excerpted abstract follows.
Friday, October 9, 2026
Power and Sample Size: The Under-Appreciated Influence of "Noisy" Data
While problems with under-powered studies are well-understood, one under-appreciated aspect involves data noise. This subtle wrinkle is discussed in a recent (and helpful) post by Andrew Gelman (click here). Gelman's main point is that, when it comes to addressing under-powered studies, "the real problem is noise, not N."
Gelman's point is well-taken, if oft-ignored. For many researchers, the go-to remedy for under-powered studies is to simply increase sample size (N). Gelman's post underscores that "[n]oise (variation in effect size, variation in outcomes, and measurement error) is more of an issue than sample size because lots of bad studies are so noisy that, to get a large enough sample size to detect a reliable effect, you’d need N = a zillion, and at that point you’d be averaging over such a wide range of conditions that it’s not clear what you’re estimating anyway." Thus, possible solutions to problems attributable to under-powered studies benefit from multivariate rather than univariate approaches.
