3dttest++ vectors warning

AFNI Message Board
1

Hello,

I was wondering what the " XXXX vectors are constant" warning in 3dttest++ indicates? I’ve seen if before, but cannot find documentation on what exactly it means or if it is a serious issue over a certain number.

Thank you!
Lea

2

Hi Lea,

It suggests that all subjects are masked (constant zero)
at those voxels. If you are not specifying a mask to
3dttest++ (which is fine), it is common to see that warning.

  • rick
3

Hi Rick,

Ok awesome, thank you so much!

Lea

5

Dear Rick,

I have a follow-up question regarding this warning. I am using a grey matter mask and I am getting

*+ WARNING: 3dttest++ -setA :: 2410 vectors are constant

How can I be sure that nothing is wrong with my mask or my data and that the warning really is nothing to worry about?

Thanks!
Marla

6

To be slightly more precise: two different kinds of things are collected into "vectors" in 3dttest++, and these vectors get checked for the error of being constant (having the same number in every entry):

  1. All the data from a single voxel across all the -setA or -setB datasets -- the "constant vectors" message means that some voxels were detected where all the numbers in that voxel were the same value -- which is not something that can be t-tested. This type of error leads to the messages shown in this thread. The most common condition is that all the dataset values from -setA (or -setB) were 0 in an offending voxel.

  2. The covariate values (from a table or from a dataset) are also collected into vectors -- and these are also checked. If all covariate values are the same, this also leads to a similar error message, but about "-setA covariate #1" (for example).

With a little effort, a 3D dataset mask of such crimes could be output to help with determining where these problems are occurring. However, since this type of error can occur in several places, you might end up with several output masks (in particular, when using the -brickwise option).