On Mon, 24 Sep 2007, Sven Schreiber wrote:
Ok, here are some issues with the GMM dialog:
First what I view as bugs:
1) choosing the "iterated" option doesn't seem to be honored, I get
I believe that was a "cosmetic" issue: the iteration option was in
fact honored, but if the thing converged immediately the printout
said "one-step". I think this should now be fixed.
2) the convergence tolerance setting is shown in the GMM output
but not anymore after saving to the session window and reopening (bad).
OK, will think about that.
3) unrelated to GMM: horizontally resizing the session window leads
icons dis- and/or reappearing
Hmm. Anyone else seen that? I tried on my own machine and
couldn't replicate the problem. It sounds like a GTK issue: is it
specific to Windows?
And some small wishes relating to the GMM GUI:
4) An option to set the tolerance would be nice (simple entry field or
even drop-down list with a range of digits maybe)
5) A checkbox for requesting the $vcv; maybe also an entry field to
assign a name to the matrix that would be created in the session.
And finally, for Gmm in general (script and GUI use):
6) Ability to specify standard options for the (initial) weighting
matrix. AFAICS that would be a choice of two: the identity matrix or the
inverse covariance matrix of the instruments. Of course over-rideable by
an explicit weights command.
These all seem like desirable things. It's just a small matter of
how many hours there are in the day.