Dear Anderson,
Thanks very much for your advice, it's much appreciated. I've tested the analysis on a subsample of my data and seem to generate some sensible looking output.
I did have a few follow-up questions:
1) I used values 0, 6 and 15 in the (Time,controls) and (Time, patients) columns, though was wondering if having a 0 would mean this gets data ignored, as there are a lot of other 0s in each column for the subjects from the other group? If that's the case, I'll try and use age in months instead.
2) For the contrasts, I used [1 0 0 0 0 0] and [-1 0 0 0 0 0] to look for increase and decrease in controls over time, with [0 1 0 0 0 0] [0 -1 0 0 0 0] for patients. But how can I look for group x time interactions? Is that possible with this approach?
3) In response to Paul Chou's reply to your message, I presume it's correct to use both the --permuteBlocks option and the "-e" option with to specify the design.grp file, not just one or the other?
Many thanks once again,
James
|