Hello,
I'm interested in using NBS to test a correlation of functional connectivity with a particular behavioral variable, controlling for other nuisance regressors. I've specified my contrast and design matrix correctly, but I'm running into issues when trying to load exchange blocks for permutation testing.
My dataset contains both sibling pairs and unrelated participants, and I would like the permutations to respect the family structure in the data such that siblings' labels are only permuted between each other while unrelated participants' labels are allowed to be freely exchanged. I've set up my exchange blocks to look something like this:
[1;1;2;2;3;3;4;5;5;6;6;7;7;8;8;9;9;10;10;4;4;4;4;4;11;11;...]
Block four contains unrelated participants, and all other blocks describe sibling pairs. Whenever I try to load a file with the exchange blocks, it isn't recognized by NBS. I've tried loading the blocks as .txt and .mat files, as well as a Matlab expression with no success. As a test, I entered a .mat file with a vector of ones with the same number of participants as my datset, and NBS recognized that file, so I'm fairly certain that I'm using the correct syntax.
Is there an apparent issue with the way I've specified up my exchange blocks? Or, are exchange blocks not suitable for this scenario? Any help would be appreciated.
Thanks,
Tyler
Hi Tyler
NBS requires the same number of observations per block. So hybrid designs including related and unrelated individuals would not work.
E.g.
[1,1,2,2,3,3] is ok but [1,1,1,2,2,3,3] is not because one block comrpsises 3 observations and the others only comprise 2. All blocks needs to have the same number.
It shouldn't be too hard to adjust the code (NBS.glm.m) to generalize the exchange blocks.
Best,
Andrew
Originally posted by Tyler Sassenberg:
Hello,
I'm interested in using NBS to test a correlation of functional connectivity with a particular behavioral variable, controlling for other nuisance regressors. I've specified my contrast and design matrix correctly, but I'm running into issues when trying to load exchange blocks for permutation testing.
My dataset contains both sibling pairs and unrelated participants, and I would like the permutations to respect the family structure in the data such that siblings' labels are only permuted between each other while unrelated participants' labels are allowed to be freely exchanged. I've set up my exchange blocks to look something like this:
[1;1;2;2;3;3;4;5;5;6;6;7;7;8;8;9;9;10;10;4;4;4;4;4;11;11;...]
Block four contains unrelated participants, and all other blocks describe sibling pairs. Whenever I try to load a file with the exchange blocks, it isn't recognized by NBS. I've tried loading the blocks as .txt and .mat files, as well as a Matlab expression with no success. As a test, I entered a .mat file with a vector of ones with the same number of participants as my datset, and NBS recognized that file, so I'm fairly certain that I'm using the correct syntax.
Is there an apparent issue with the way I've specified up my exchange blocks? Or, are exchange blocks not suitable for this scenario? Any help would be appreciated.
Thanks,
Tyler