First column of 1s in the design matrix for connectomestats

[orencivier]

Hello @rsmith,

To me, it seems that your the response here (suggesting that a first column of 1s is not ideal):

Question about 'design matrix' and results of 'fwe_pvalue' when using connectomestats nbse

(1) I am not sure if the design matrix and contrast vector I wrote is correct. And would you like to check them for me. Your design matrix is not ideal, as it is rank-deficient: the first column can be represented as the sum of the second and third columns, which means that there are an infinite number of solutions to the model fit. You want to either have one column per group, and extract the difference between them: 1 0 1 0 0 1 0 1 1 -1 , or capture the population mean in one col…

Contradicts the documentation for connectomestats (claiming that “In some software packages, a column of ones is automatically added to the GLM design matrix”):
https://userdocs.mrtrix.org/en/latest/reference/commands/connectomestats.html

Or is it that a simple design matrix of

1  0
1  0
0  1
0  1

is one of “rare situations where including such a column would not be appropriate for a particular experimental design”. It’s quite confusing, because I wouldn’t say that such a trivial design matrix is rare. On the opposite, it is probably the most common one being used.

I apologise if I misunderstood something, but I think a clarification is warranted regardless.

Thanks ahead,
Oren