3dttest++ VS 3dDecon comparisons

AFNI Message Board
1

Hello,

I am interested in comparing two different stimuli classes wordsVhouses and found that when I look at the output for when I have 3dDeconvolve do the comparison and the output for when I do a two sided t-test in ttest++ they are different.

In the first case I compared wordsVhouses as a GLT in my 3dDecon script and then put that subbrick into 3dttest++ as a one sided t-test. In the second case I took the subbrick for words and the subbrick for houses and compared the two in ttest++. The results are similar, but the 3dDecon is noisier and I am wondering why that is, should these be doing something very similar?

Thank you for your help!

2

The two approaches should render virtually the same result except for rounding errors. In addition, the issue of one- vs. two-sided testing is not relevant; that should be the case at least in AFNI as well as in statistics in general except that some software packages may have a peculiar setup.

3

Be sure you are doing a paired test.

  • rick
4

That was the issue… I had left off the -paired option, ugh such a little thing can cause so much grief! Thank you so much!