Hi -

I'm helping my son with his science fair project. He did placebo taste testing where he has some blind taste likert data (on a scale of 7) as well as some "informed" taste likert data. He did the test six ways: A-B (blind), B-A (blind), A-B (accurate info), A-A (inaccurate info), B-A (inaccurate info), and B-B (inaccurate info).

Sample:
TestType Item1 Rating1 Item2 Rating2 Blind
1 A 5 B 3 1
2 A 6 B 7 0
3 A 4 A 4 0

We are going to do an ordered logit by separating each half of the test and adding variables that capture the rest of the information - blind, A, B, first, second, etc.

But I think that's not going to cut it because it's the relative rankings we are interested in: How does the info change the flavor tests? Is there a way to handle this in stata? My gut says that we can normalize by taking the mean of the differences of the blind tests, and then seeing how the mean of the differences differs in each of the treatment groups. Is there a special test for that, or is it just standard parametric/non-parametric testing?

Any thoughts appreciated - this doesn't have to pass peer review, but he would like to do the best analysis he can, and I'd like him to use Stata to make it easier (and to learn it)!