Hi Harry,
Have you played around with creating actual contrasts within the GUI?
I haven't been able to get Matlab launched on my MacServer today to
check the actual message provided, but at least in SPM8 the GUI for
setting up contrasts prevented the definition of non-estimable
contrasts, and if I recall correctly, the message it provided was
something along the lines of "invalid contrast". I don't recall it
providing explicit hints/suggestions as to what would make the contrast
valid.
cheers,
-MH
At least within SPM8
On Thu, 2010-07-29 at 10:52 -0400, Harry Graber wrote:
> There are theoretical issues that I’m confused about: one of my quirks
> is I have serious anxiety about using any method until I understand it
> well enough to explain/teach it to other people, and until I can re-
> derive the principal formulas on my own if necessary. The questions I
> am concerned with here grow out of my readings in the general linear
> model chapter (www.fil.ion.ucl.ac.uk/spm/doc/books/hbf2/pdfs/Ch7.pdf)
> of Human Brain Function, 2nd Ed.
>
>
>
> 1. The section about estimable functions and contrasts (p. 9) was
> slow-going, but I’m glad I stuck with it. The material I had read
> before (e.g., www.utdallas.edu/~herve/abdi-contrasts2010-pretty.pdf)
> just said that a linear combination of data groups is a contrast if
> the mean of the coefficients is zero. According to Ch7, which I’m
> inclined to trust more, a vector c is a contrast vector if it a linear
> combination of the rows of the design matrix X and is invariant under
> re-parameterization of the model. Mathematically, the second
> criterion is that c is unchanged when post-multiplied by (MATLAB
> notation) pinv(X’X)X’X.
>
> It occurs to me that if X’X is invertible, then any linear
> combination of rows of X is a contrast. True? (I recognize that X’X
> probably is non-invertible most of the time, in practice)
>
> It is stated in Ch7 that SPM tests every user-specified contrast
> to determine whether or not it really is a legitimate contrast. But
> what does SPM do if the answer is “no”? Notify the user in some way?
> Give the user a chance to try again? Offer some sort of guidance in
> specifying a contrast that is proper? Automatically “repair” the
> invalid contrast in some way (with or without notifying the user of
> the fact)?
>
> My firm impression after reading the technical description of
> contrasts is that sum(c) = 0 is neither a sufficient nor a necessary
> condition. However, at the top of the very next page (p. 10), the
> authors of Ch7 acknowledge that “For most designs, contrasts have
> weights that sum to zero over the levels of each factor.” Does SPM do
> anything to force this property upon user-specified contrasts?
>
> I think it prudent to leave off here, and not let this e-mail
> grow until it needs to be rolled up on a stick. The rest can wait.
> Thank you for any answers you can provide.
>
>
> Harry L. Graber, PhD
> Research Assistant Professor
> Department of Pathology
> SUNY Downstate Medical Center
> 450 Clarkson Avenue
> Brooklyn, NY 11203
>
> (718) 270-1286 (phone)
> (718) 613-8774 (fax)
> http://OTG.downstate.edu
>
|