help
help > RE: NBS Design matrix for EEG connectivity
Dec 1, 2021 10:12 PM | Andrew Zalesky
RE: NBS Design matrix for EEG connectivity
Hi Tyler,
thanks for sending through your design matrix.
The design matrix looks fine. The contrasts also look fine.
I suggest also trying the contrast: [0,0,0,0,0,0,0,0,0,0,0,0,-1]. Note the "minus" in front of 1. This will test for reduced connectivity in the first group, whereas without the minus will test for reduced connectivity in the second group. Let me know how this goes.
Make sure that you select "t-test" from the drop down menu.
p-values of exactly 1 seem suspicious, but they are possible.
best,
Andrew
Originally posted by Tyler Durden:
thanks for sending through your design matrix.
The design matrix looks fine. The contrasts also look fine.
I suggest also trying the contrast: [0,0,0,0,0,0,0,0,0,0,0,0,-1]. Note the "minus" in front of 1. This will test for reduced connectivity in the first group, whereas without the minus will test for reduced connectivity in the second group. Let me know how this goes.
Make sure that you select "t-test" from the drop down menu.
p-values of exactly 1 seem suspicious, but they are possible.
best,
Andrew
Originally posted by Tyler Durden:
Hi Andrew,
thank you for the quick reply and your help. Indeed I did not see the help option. I have constructed my data according to your comments: So I have concatenated connectivity matrices (pre & post per participant) for 12 participants and the resulting connectivity matrix has the dimensions 32x32x24. I used the paired t-test as a statistical model. However, when I run NBS with 12 participants, all p-values are 1.0. I then tried to run the NBS with less participants (n=3) and it works there or at least I get more regular p-values. I attached my design matrices for 12 participants because I think there might be a mistake in there...although I built it according to the help function. The contrasts I used were [0,0,0,0,0,0,0,0,0,0,0,0,1] for n=12 and [0,0,0,1] for n=3.
I'm fairly new to this so please excuse these basic questions.
Best,
T.
Originally posted by Andrew Zalesky:
thank you for the quick reply and your help. Indeed I did not see the help option. I have constructed my data according to your comments: So I have concatenated connectivity matrices (pre & post per participant) for 12 participants and the resulting connectivity matrix has the dimensions 32x32x24. I used the paired t-test as a statistical model. However, when I run NBS with 12 participants, all p-values are 1.0. I then tried to run the NBS with less participants (n=3) and it works there or at least I get more regular p-values. I attached my design matrices for 12 participants because I think there might be a mistake in there...although I built it according to the help function. The contrasts I used were [0,0,0,0,0,0,0,0,0,0,0,0,1] for n=12 and [0,0,0,1] for n=3.
I'm fairly new to this so please excuse these basic questions.
Best,
T.
Originally posted by Andrew Zalesky:
Hi Tyler,
take a look at the NBS Manual or click on the Help button. Numerous design matrix examples are provided.
For you experiment, you would perform a repeated measures t-test. It is repeated measures because the two measurements are acquired for each subject (i.e., pre and post).
The concatenated connectivity matrices should be of dimensions 32 x 32 x N, where N is double the number of subjects. Double because each subject has a pre and post connectivity matrix.
I am not sure why the third dimension is 2 in your case.
The number of rows in the design matrix must equal N.
best,
Andrew
Originally posted by Tyler Durden:
take a look at the NBS Manual or click on the Help button. Numerous design matrix examples are provided.
For you experiment, you would perform a repeated measures t-test. It is repeated measures because the two measurements are acquired for each subject (i.e., pre and post).
The concatenated connectivity matrices should be of dimensions 32 x 32 x N, where N is double the number of subjects. Double because each subject has a pre and post connectivity matrix.
I am not sure why the third dimension is 2 in your case.
The number of rows in the design matrix must equal N.
best,
Andrew
Originally posted by Tyler Durden:
Dear all,
I have two connectivity matrices (32 channel EEG) holding average phase lag indices of Condition A (pre) and Condition B (post) and I aim to compare them using the network based statistics toolbox (NBS). The concatenated connectivity matrix has the dimension 32x32x2. I also loaded the node coordinates. However I have trouble understanding the design matrix. What does the matrix need to look like if I want to obtain the statistical differences between both matrices? I tried some example but I always get the error: "Design matrix not found or inconsistent". I hope you can help me!
Best,
T.
I have two connectivity matrices (32 channel EEG) holding average phase lag indices of Condition A (pre) and Condition B (post) and I aim to compare them using the network based statistics toolbox (NBS). The concatenated connectivity matrix has the dimension 32x32x2. I also loaded the node coordinates. However I have trouble understanding the design matrix. What does the matrix need to look like if I want to obtain the statistical differences between both matrices? I tried some example but I always get the error: "Design matrix not found or inconsistent". I hope you can help me!
Best,
T.
Threaded View
| Title | Author | Date |
|---|---|---|
| Tyler Durden | Nov 26, 2021 | |
| Andrew Zalesky | Nov 26, 2021 | |
| Tyler Durden | Nov 30, 2021 | |
| Andrew Zalesky | Dec 1, 2021 | |
| Tyler Durden | Dec 7, 2021 | |