Hi,
We are using a program called Fidap to do multiple regression on
neuroimaging data. The problem is nobody here or where the program came
from understands how to select coefficients, ie., whether and how to make
them orthogonal. I am not very sophisticated statistically, but can muddle
through a book if it's not just equations. Would anyone recommend an
understandable book on multiple regression? Thanks very much for your
help.
I'll explain the problem a bit in case it helps in understanding what I
need to learn.
We administer problems of type A or B. Each has two main periods: solve
and rest. We want to find areas in the brain that respond to a particular
epoch during each type of problem. So I guess the factors (regressors?)
would look like (though they would be extended over all trials):
A solve | rest B solve | rest .......
|~~~~~~~~~~~|____________________________________________________
____________|~~~~~~~~~~~|________________________________________
________________________|~~~~~~~~~~~|____________________________
____________________________________|~~~~~~~~~~~|________________
So we select coefficients for these regressors' epochs as such:
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
We do not know whether these four regressors (can we call them vectors?)
need to be orthogonal. Especially as we get into more complicated analyses
where the regressors or effects overlap or coincide. Nobody who uses the
software we're using has used more than two very simple regressors so
nobody is able to advise us, so we need to learn about multiple regression
and figure out what to do ourselves.
Thanks much.
Jim
--
Remove SPAMBLOCK to reply
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
