[S] Vectorized Calculations & Optimization

Timothy Johnson (tjohnson@s.psych.uiuc.edu)
Fri, 17 Apr 1998 10:24:01 -0500 (CDT)

Here's something that I have had trouble with recently - using
functions ms or nlmin where I am minimizing a negative log likelihood
and I need to incorporate vectorized calculations of some sort. For
example, in section 9.7 of V&R (2nd edition) there is an example of
fitting a logistic regression with ms. Here, the function is
relatively simple. But, with a more complicated design it would be
much easier to use vectorized calculations with a parameter vector
and design matrix to specify the model. Of course, we don't have to
do this for logistic regression because we have glm but I'm having to
fit models that have somewhat complicated design matrices and are not
of your typical glm sort.

I have not had any luck using vectorized calculations or a similar
shortcut with ms or nlmin. One solution I have used is just writing
the entire function in terms of dummy variables but that is tedious.
For example, one of my models is similar to a Rasch model so it
involves a lot of unknown parameters but it isn't one such that I can
simpify the fitting process by conditioning on the total score as is
normally done with true Rasch models.

Thanks for your help.

Tim Johnson
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news