 |
 |
If you were going to combine the 4 trials into one regressor, am I right in thinking that it would be regressor 1 + regressor 2 + regressor 3 - regressor 4? Then the estimate is simply whether the first three are greater than the last one. Or would regressor 4 need to multiplied by 3 first (e.g. the amplitude of the 4th trial is 3 times as great as the other ones in the model). I don't think the latter is correct because if there were 2 regressors (1+2+3 and 4), the contrast [1 -1] would give you the wrong solution if you first multiplied by regressor 4 by 3 (e.g. the beta would 1/3 of the others even it was the same response). If these are correct, then collapsing regressors should be done without weighting them. On Tue, Jun 2, 2009 at 6:30 AM, Karl Friston <[log in to unmask]>wrote: > Dear Donald, > > Perhaps incorrectly, I've been thinking of interactions in the following > way: When interactions are formed, dummy codes are not typically centered > first (e.g. grouping dummy variable*(centered continuous variable) and the > resultant term represents the additive effect of the continuous variable in > the group that is not set to 0. Centering a grouping dummy variable would > depend on the size of the group relative to the other group. To me, any > procedure should be independent of group size or in the case of PPI, number > of trials. > > Since, psycho-physiological interactions are using dummy codes or their > linear combination, should they still be centered? > > Any thoughts would be appreciated. > > > > Yes, generally all explanatory variables (that use dummy or indicator > variables or not) should be > mean centered when forming interaction terms (this is not special to PPIs > but it implicit in > any ANOVA or ANCOVA) - one does not usually need to do it explicitly > because the main effect > and mean (constant) terms are part of the model and are therefore explained > away when testing > for the interaction* per se*. > > In the case of an unequal number of trails it may help to consider a test > for a difference in activation between > an 3 trials of one sort and 1 trial of another with a regressor [1 1 1 -1]. > However, what you are actually testing > is the effect of this regressor that cannot be explained by the constant > terms [1 1 1 1]. This effect is > [1 1 1 -3]. If it was not then you would get a significant effect of the > difference if there was no difference > between the trials; because there are more of the first trial types. > > I hope this helps - Karl > > -- Best Regards, Donald McLaren ===================== D.G. McLaren University of Wisconsin - Madison Neuroscience Training Program Office: (608) 265-9672 Lab: (608) 256-1901 ext 12914 ===================== This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is intended only for the use of the individual or entity named above. If the reader of the e-mail is not the intended recipient or the employee or agent responsible for delivering it to the intended recipient, you are hereby notified that you are in possession of confidential and privileged information. Any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited and may be unlawful. If you have received this e-mail unintentionally, please immediately notify the sender via telephone at (608) 265-9672 or email.