 |
 |
See below. On Sun, Jan 19, 2014 at 5:49 AM, Reut Moran <[log in to unmask]> wrote: > Dear experts > > I have some questions with regard to covariate treatment in the > statistical model and multiple regressions: > I would like to find out if the covariate "age" has an influence in my > experiment. How do I use a two-sample t-test with age as a covariate? > Can I do it without using a full factorial design? In which I define 2 > factors: group and age. > Yes. You use two-sample t-test with a covariate. > > Another question is about multiple regressions; we used a design matrix of > multiple regressions in order to find out whether protein levels influence > the brain activity. We used 3 contrasts: > Group: 1 0 > Positive: 0 1 > Negative: 0 -1 > Is that the right way to do it? You need a constant term in the model. It may be easier to use a two-sample t-test, in which the 3 columns would be: group1, group2, protein levels. Group: 1 -1 0 Positive: 0 0 1 Negative: 0 0 1 I'd read the following site on how to interpret the group effects and whether you should mean center the results: *mumford*.fmripower.org/*mean*_ *centering*/ > Also, we can use some help in interpretation of the results: Would it be > correct to say that regions that are negatively correlated with the protein > level, means that these areas have decreased activity the more protein > exist? And visa verse for areas positively correlated with the protein > level? > Yes. But it could also be interpreted as: (1) negative -- more activity with less protein; and (2) positive -- more activity with more protein. I'd avoid the word decrease and increase because you haven't said what your comparison is. > Finally, the results show that there is laterality in activation, meaning > that there is a positive correlation between the left hemisphere and > protein level, and negative correlation to the right hemisphere. Looks > weird. Can these be due to a mistake in our method? > Perhaps or it could be different in each hemisphere. > > > Thank you in advance for your help. > Reut >