Hello,
I have a study with three parametric regressors (X, Y, and Z). In addition to looking at the main effects of all 3, I am also interested in the interaction two of them (X, and Y). To model this, I created four event files (3 columns, the first with just ones in the third column, and then the other three with the variables of interest). To create the interaction term, I used the interaction button in feat, and selected columns 2 and 3. I selected 'mean' as the way of computing the interaction.
All the main effects and interactions that I would expect seem to have come out, but now I would like to plot the expected scores at different levels of my continuous variables - and this is where I have become confused.
Suppose I wanted to know the value at X = 2, Y = -1, and Z = 0, I would think that I could just take do something like:
2*PE2 + -1*PE3 + 0*PE4 + (2*-1)*PE5.
When I look at the matrix, all the variables mean to 0, but when I multiply the columns, they are not what you would expect through multiplications.
Can someone give me advice about how to compute expected scores? I guess what this comes down to is understanding how the raw variables from the 3 column input files are converted into the design matrix so that I can reverse the process for the plot.
Thanks.
Wil
p.s., This also gets more complicated when I want to look at X^2 (which I created a new EV for). Those values do not sum to zero (all are positive), but I hope that I can figure out how to deal with these with the answer to above. But, if that is different, please help.
|