CODEWITHSUNDEEP
statisticswekasignificance
weidler
weidler

Reputation: 676

Paired T Test for different variable sets

Does it make sense to compare the results of a machine learning algorithm for different variable/attribute sets but the same Sample?

I am comparing different variable/attribute groups to find the most efficient ones. Normally you use t-tests to compare if another sample has significantly different results on the same variables. Can I do it vice versa?

Upvotes: 0

Views: 155

Answers (1)

JSN
JSN

Reputation: 451

If you want to compare two values derived from one sample you can use an one-sample t-test.

However, if you want to compare more than two values you shouldn't use t-tests due to the multiple comparisons problem. Basically this just means that the more t-tests you calculate the higher is the chance to get a type 1 error. In order to avoid that you could calculate a repeated measures ANOVA with pairwise comparisons (e.g. Bonferroni).

Upvotes: 1

Related Questions