Errata and Responses

(back to the main Bad Stats page)

Corrections to ``HCI Statistics without p-values'' (research report)

  1. The p-value for Pill 4 is .012, not .001 (Figures 2, 3, 7 and Table 1). Fortunately this does not change the argument, and even makes it slightly stronger.
    (Added 25 June 2015) Thanks to Geoff Cumming for spotting this.
  2. I revised my definition of estimation, and believe it should refer to reporting effect sizes with interval estimates in general, rather than confidence intervals specifically. Confidence intervals are only one type of interval estimate, a closely related one is the Bayesian (credible) interval. Credible intervals are the only intervals for which interpretations such as "a range of plausible values" are formally correct. They have many advantages but are harder to master and still not widely used and supported by stats packages. For now, my position is that moving from dichotomous testing to estimation is a larger step forward than moving from frequentist to Bayesian calculation procedures (many would agree, many would also disagree). In typical HCI analyses, which are generally quite simple, confidence intervals often equal or approximate credible intervals with uninformed priors and are thus a perfectly good place to start. HCI researchers who are perfectionists and are willing to invest time can of course directly switch to proper Bayesian estimation, but it is also acceptable and natural to switch only after having gained some experience with estimation using confidence intervals, and/or after Bayesian methods become more firmly established in research practice.
    (Added 25 June 2015, updated 8 July 2015) Thanks to Dan Simmons and Steven Franconeri for providing me with thought-provoking reading material on this.

This is all addressed in the final revision of the chapter. This revision discusses Bayesian intervals vs. CIs (with refs), clarifies the differences between Fisher and Neyman-Pearson's approaches to statistical inference, and recalls the original definition of confidence intervals by Neyman. Also, some tips have been reworded and clarified based on comments from the editors (Maurits and Judy) and others. Download author version.

Corrections to ``Fair Statistical Communication in HCI'' (book chapter)

  1. At least four times in the chapter, I use the term exact confidence interval to refer to the confidence interval based on Student's t-distribution. While it is indeed an exact confidence interval, there are many other types of exact confidence intervals. It would have been preferable to use the term Student t confidence interval or t-based confidence interval.
    (Added 21 September 2017)
  2. I wrote that in many common situations, confidence intervals agree with so-called objective credible intervals. It would have been perhaps clearer to state that in many common situations, confidence intervals agree with credible intervals obtained using flat or non-informative priors.
    (Added 21 September 2017)
  3. Whenever I refer to bootstrapping, I actually mean non-parametric bootstrapping more specifically. Parametric bootstrapping is a different and more complex approach which assumes we have fairly good knowledge of the population distribution.
    (Added 29 September 2017)

New and updated tips

Nothing else for now, but I do keep some notes and may add new tips here at some point.


I'm happy to publish questions, comments, commentaries and critiques about the final chapter revision here. (none received for now)


All material on this page is CC-BY-SA. You can reuse or adapt it provided you link to