SPM users,
We have completed an event-related design where events of interest are
separated in time by a minimum of 12 seconds. Our interest is in looking
at each subject's responses to stimuli as the basis for defining conditions
in the design matrix (for example, errors committed to different types of
stimuli). To do this, each subject's particular pattern of responses to
the stimuli forms the basis for their own design matrix at the first level
of analysis. The problem is how to handle subjects who HAVE NOT produced a
particular type of response when other subjects HAVE (for example, errors
to a particular type of stimulus). We would like to explicitly model the
particular response as a distinct event type for those subjects who
produced it so that we can assess its associated activations in a random
effects model involving those subjects. Obviously, we cannot explicitly
model this distinct event type in subjects who did not produce the defining
response, but would still like to use all of the subjects in random effects
models examining response types exhibited by ALL subjects.
The question can be posed as follows:
-We apply a design matrix to some subjects that includes the explicit
modeling of event type A (a particular response to a stimulus type) as well
as event types B, C, and D.
-In other subjects, the design matrix only includes event types B, C, and D
because event type A did not occur.
Can we legitimately do a random effects model comparing event types B and C
using all subjects, ignoring the fact that the design matrix included event
type A in only a subset of subjects?
Alternatively, should we only include in the design matrix those event
types that occur in all subjects, relegating any remaining event types that
occur in only a subset of subjects to the residual variance (ie, implicit
baseline)?
Daniel H. Mathalon, Ph.D., M.D.
Assistant Professor
Department of Psychiatry
Yale University School of Medicine
Psychiatry Service 116A
VA Healthcare System
950 Campbell Avenue
West Haven, CT 06516
Phone (203) 932-5711, ext. 5539
FAX : (203) 937-3886
Pager 203-867-7756
e-mail: [log in to unmask]
|