 |
 |
Hello, I am trying to make sure I am doing a 2-group t-test with additional variables (age and sex are tipical choices) correctly, according to the instructions at www.fmrib.ox.ac.uk/fsl/randomise/index.html. For space/illustration purposes, let's say I have a group A of N_A=3 subjects, and a group B of N_B=4. The design.mat/con files produced with the design_ttest2 tool and used in the randomise command are: ==== design.mat with no additional variables: /NumWaves 2 /NumPoints 7 /PPheights 1 1 /Matrix 1 0 1 0 1 0 0 1 0 1 0 1 0 1 ==== design.con : /NumWaves 2 /NumContrasts 2 /PPheights 1 1 /Matrix 1 -1 -1 1 ==== ==== ==== My understanding, should I wish to add an additional (confounding) variable, like age, is that I should add another column (with appropriate age values) to the design.mat file, and leave the contrast file and everything else unchanged (thereby the contrast ignoring the new variables): ==== design.mat with additional -age- variable: /NumWaves 2 /NumPoints 7 /PPheights 1 1 /Matrix 1 0 24 1 0 48 1 0 35 0 1 29 0 1 47 0 1 22 0 1 31 ==== ==== ==== and that I should use exactly the same randomise command as before to compute the tfce_p_tstat images, with the obvious modification should I wish to include even more (confounding) variables. Is this correct? Also, when the simple 2-group t-test is done, a number of (N_A + N_B)!/( N_A! * N_B! ) maximum permutations are possible to choose N_A subjects for the first group. However, when additional variables are used, there seems to be room for more shuffling rather than just group membership. Actually, I ran the randomise command with the -q option for these design files and I obtained an exhaustive number of permutations equal to 2520=35*72 in place of the 35 initial ones (before adding the 3rd column). Is there a simple explanation for this number (the multiplicative 72 factor)? Thank you very much, Silviu Podariu UNMC, Omaha