Hello,
I am attempting to run a linear regression that would look at the relationship between the cognitive score and the connectivity matrices of 22 subjects, controlling for age and gender. When I run, I produce an output that visualizes the correct connections and tells me that p<.0001. I noticed, however, when I change the connectivity matrix, I still get the exact same output with the same p value.
Contrast matrices tried:

• Initially I did not have an intercept
  [-1, 0, 0]
  [1, 0, 0]
• I then added an intercept, and changed the contrast value, which had no effect on the result
  [0, -1, 0, 0]
  [0, 1, 0, 0]
  [1, -1, 0, 0]
  [1, 1, 0, 0]
• I even tried using one of the values I wanted to stay constant as a predictor by switching it over to 1- no effect
• Tried it on another machine
• Lastly, I made a text file with the connectivity matrix instead of typing it directly into the GUI, but that too made no difference.

I'm fresh out of ideas and would be grateful for any assistance.
Best,
Jonika

Hi Jonika,
It might be helpful if you provide details about your design matrix and connectivity matrices.

Note that you will need a connectivity matrix for each subject. When you say that you changed the connectivity matrix, it is not clear which subject. In any case, if you change the connectivity matrix for only one subject, this will probably not have a substantial effect on the results. So it should not be surprising that the results remain unchanged when you only change 1 of the 22 connectivity matrices. 

Andrew

Originally posted by Jonika Tannous:

Thank you Andrew for your response.
I misspoke/misstyped: I meant that I changed the contrast matrix, not the connectivity matrices. I originally wanted to do a linear regression that would look for a negative relationship between the cognitive score and the connectivity matrix while controlling for sex and age. I am using the GUI and have the statistical test on one-sample. I have 22 16X16 connectivity matrices, all text files within 1 folder. My design matrix is 22X4 with the columns representing the intercept, cognitive score, sex, and age respectively. The format looks like this:

[1     62.5       1      48.4
1      30.56     1      52.5
.....
1      20.83     1      41.6]

The problem arises with the contrast matrix: I used a contrast matrix of [1, -1, 0, 0]. This gives me a valid-looking out put with a p-value <.0001. When I change the contrast matrix, say to look for a positive relationship, to [1, 1, 0, 0], I get the exact same output, same p-value, and the same t stat at each edge. Changing the contrast matrix in many other ways (refer to the original message), still produces the same output.

Thanks,
Jonika

Originally posted by Andrew Zalesky: