Greig,
>
> Dear SPMers,
>
> I have a question relating to the reporting of SPM{F}s.
>
> I have data from a parametric study in which the difficulty level of a
> particular task was linearly incremented over 3 conditions with an
> additional appropriate baseline condition interspersed among them. I
> employed a single linear contrast to represent this with SPM97. In this
> case, is it appropriate to report the SPM{F} rather than the SPM{T}/{Z}?
>
> thanks,
>
> Greig de Zubicaray
Do you mean single, linear covariate as opposed to "single linear
contrast"? If so, that means that you have only one independent variable
representing all of the conditions. In such a case, the SPM{F} and the
SPM{t} will yield equivalent significances (and are testing the same
thing).
If however you mean that you reference coded the conditions to baseline
then you will have three covariates representing the condition effects
with respect to baseline (if this particular reference is chosen). In such
as case, there is no way (at least not to my understanding) to equate the
meaning of a contrast (which is tested via an SPM{t}/{SPMz)} to that
of the condition effects taken in ensemble (which is tested via an
SPM{F}). In other words, the correct choice would depend on what one is
testing: a linear compound of parameter estimates or the total
explanatory power of the covariates of interest.
Sincerely,
Eric Zarahn
University of Pennsylvania
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|