[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: Fixing an -ml model- syntax problem
At 12:45 PM 2/20/2007, Arne Risa Hole wrote:
In Allison's paper, he almost makes it sound like the fact that
estimated error variances can go negative is a feature, not a bug,
i.e. negative error variances suggest that the model makes no
sense. My own experience with his code is that it doesn't actually
produce these impossible estimates under the conditions he describes,
it just refuses to converge and produce any estimates at all. In
general I've found that oglm's coding tends to work faster and more
reliably than Allison's coding.
forces the estimated scale parameter/ error variance to be positive,
which is a good thing. This is a similar parameterisation to that used
in Richard Williams' -oglm- routine.
Incidentally, the citation for Allison's aper is
Allison, Paul. 1999. "Comparing Logit and Probit Coefficients
Across Groups." Sociological Methods and Research 28(2): 186-208.
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
WWW (personal): http://www.nd.edu/~rwilliam
WWW (department): http://www.nd.edu/~soc
* For searches and help try: