I just came across the paper below by White & Gorard who argue against the use of inferential statistical tests:

"the assumptions underlying inferential statistical tests are rarely met, meaning that students are being
taught analyses that should only be used very rarely. Secondly, all of the most
common outputs of inferential statistical tests – p-values, standard errors and
confidence intervals – suffer from a similar logical problem that renders them at best
useless and at worst misleading. Eliminating inferential statistical tests from statistics
teaching (and practice) would avoid the creation of another generation of
researchers who either do not understand, or knowingly misuse, these techniques."

As an instructor and statistical consultant, who apparently does not understand and has unknowingly misused these techniques, I would be very interested to hear the thoughts/opinions of others on this topic!




Not a robit
Well the NHST dilemma is always an issue along with why 0.05 and people's interpretation of frequentists approaches.

I was also thinking about the continual push by some to change 0.05 the other day and thought, why do I have to do that when I try to use regularization, corrections for false discovery, effects estimates with precion, and reasonable study designs and conclusions. But I get the issue with the significant yes/no problem which negates marginal effects and misinterpretation. That should be cleaned up.
Thanks for the reply hlsmith!

However, the question I would like an answer to is: Am I doing a disservice to my students in teaching them what I was taught? In other words, are all inferential statistical tests (t tests, ANOVA, Chi-square, etc.) fundamentally flawed? Should I discontinue teaching these things?



Not a robit
No, keep teaching - you just have to tell them to consult a statistician if ever thinking about delving into this area for real or desiring to publish generalizable results.

I get a person knocking on my door every couple of weeks wanting me to get them a pvalue. The person's study question, variable definitions and designs are completely inadequate. They don't come to me knowing that I have to completely design their study in order to get data adequate for analyses. People don't know what they don't know. But they need to ask for help or the reproducibility crisis is going to make everyone lose trust in "Research" and hopefully not "Science".
Last edited:
Statistics is a discipline in which the primary aim is to understand (and make conclusions about) a population. You cannot do this without inferential statistics. The more I think about that paper, the more I dislike the stance. I haven't read the article, but I can't see how the frequentist perspective of statistics can be taught without the central limit theorem.
Last edited: