On 23 Apr 2010, at 22:07, Qihong Zou wrote:
Dear all,
I want to ask you some questions about melodic ICA.
I am working on drug response data on anesthetic monkeys. Since we don't know the exact response curve of the drug effect on monkeys, ICA must be a perfect method for our exploratory purpose. Furthermore, as the design (onset time of drug injection) are the same across the monkeys, tensor ICA is preferred, right?
As long as they respond to the drug in similar ways to each other, yes.
Based on the components from tensor ICA, we might get several components related to the drug effect. If I go the "report" folder, I can find time courses of each component with files named begin with t*.txt. For each component, is the first column the time course for the group component and are the other columns for all individual subjects? How can we get the individual IC maps? Using dual regression approach?
Certainly one of the more straightforward ways to get individual spatial maps and timecourses is to use dual-regression, yes. If you want to try the beta-code for that send me or Christian an email and we can send you the link.
What do the reconstructed IC maps mean? Beta values (is it equal to response amplitude?) for drug activation? What maps should we used for further statistic analysis?
You can extract "beta" maps or zstat maps (etc) from the output of the dual regression - it is up to you which you use, as they allow you to ask different questions of the data.
Cheers, Steve.
Thanks a lot in advance!
--
Qihong
