@StatModeling Indeed! I had gpt4 translate some complex jags code to Stan this week and it did it perfectly after I explicitly defined each variable. Amazing time saver!
@rnsoicher I don't think it makes sense to use a paired t test since these students aren't paired. If you had samples of multiple classes, you could use a mixed effects model or something to account for within class correlation but in this case the independent test makes sense to me.
@rnsoicher Hey! I see, I guess I'd agree that the output of the t test from a statistical software is more informative. Regarding dependence/independence, are you just comparing a full class that's split into two groups? Are you assuming the class is a sample of some larger pseudopop?