Adriano Polpo, Associate Professor at UWA, spoke in the June meeting about Optimal Sample-Size-Dependent Significance Levels. The talk was based on the work (DOI: 10.1080/00031305.2018.1518268) published on the The American Statistician special issue about “Statistical Inference in the 21st Century: A World Beyond p < 0.05”. It can be seen that this is a hot topic given the number of publications in this issue is over 40. But Adriano and his coauthors address or revisit the nuances of hypothesis testing in the classical approaches in order to prepare the way for their approach to what has recently been a vexed question involving whether or not or how to use p-values?
For those who do not know Adriano, he has recently taken up his position at UWA having come from the Federal University of Sao Carlos, Brazil, where he spent 12 years. He originally got his PhD from the University of Sao Paulo in the year 2005 and has also taken post-doctoral study at Florida State University. He was born in Sao Paulo, Brazil, but also has dual citizenship with Italy. On the other hand, he does not confess to speaking Italian. Interestingly he tells me that he learnt English by using the computer.
The statisticians know of the inconsistency, or paradox, in the current classical tests of significance that are based on p-value statistics that are compared to the canonical significance levels (10%, 5%, and 1%). Adriano and his colleagues argued that researchers do not need to completely abandon the p-value, rather, they should instead stop using significance levels that do not depend on sample sizes. A testing procedure was presented, with a significance level that is a function of sample size, obtained from a generalized form of the Neyman–Pearson Lemma.
At the conclusion of his talk several people dined afterwards and continued discussions at the Bateman Chinese Malaysian Eating House.