Something went wrong in my previous message:
Mark, you said: " If what you want to do is a correlation then you either need to remove the mean from both your regressors and your data, *OR* include a regressor that is all ones (and have a 0 associated with it in all contrasts). Either of these methods will eliminate the mean value correctly from the calculations. Otherwise, the mean value will bias your results"
For regression it does not matter if the regressors are zscored or not, at least in Matlab you can test it. Correlation is how the variables covary in time, so it won't matter if the distance between them is greater or not. The corr coeff are exactly the same. In regression you need to add the column of ones, that's true.
Is this wrong? I'm very confused about this now.
Ib
|