I have a question regarding a GLM for TBSS. I would like to test the effect of a continuous variable (e.g., reaction times, RT) on FA. I also need to control for age and gender (because there are significant effects of age and gender when tested separately). Then I need to test all possible interactions: age x gender, age x RT, gender x RT and age x gender x RT.
I would like to include the following variables:
EV1: age (centered on the mean (demeaned))
EV2: gender (I set up F=1, M=0, according to the instruction in the FSL user guide, then also centered on the mean. Since the number of males (M) and females (F) is not exactly the same, the resulting values are: 0.52 for F and -0.48 for M.)
EV3: RT (centered on the mean)
EV4: age x RT (first, I calculated a product age(years) * RT(ms), then calculated the mean for these products, then centered on the mean) - is this a correct approach?
EV5: gender x RT - ? (If I proceed in a similar way as with the EV4, I will have to multiply by 0 (M=0). Would it be correct to use instead of F=1 & M=0 already centered values F=0.52 & M=-0.48?
EV6: age x gender - ? The same question, how to set up the gender?
EV7: age x gender x RT - ? Would it be correct to calculate a product of the 3 values and then center it on the mean?
Is this model correct in general?
Thank you very much for your help!
|