Dear reader,

The silence I've been experiencing on queries concerning sources for ways to 
simulate the multinomial logistic regression model (cf. my query of Nov. 16) 
strikes me as particular. Either simulating the model most be very simple or 
it is difficult. I suspect the latter. What exactly then, is this difficulty? 
The choice is either between simulating a multinomial distribution and then 
compute the model parameters; or you simulate a general linear model (i.e. 
manipulate the model parameters) and afterwards look how the resulting 
multinomial distrinution looks alike. In the latter case (which is easiest to 
implement), problems in expressing the polytomous response variable exist. In 
the first case, the mathematical form of the model is unclosed. Is there a 
third possibility? And if not, which of the two methods proposed here should 
be prefered?

With regards,

Jarl Kampen
IPSoM (Institute for Political Sociology & Methodology), Faculty of Social & 
Political Sciences, Catholic University of Brussels, Vrijheidslaan 17, 1081 
Brussels, Belgium. Voice +32(0)2 412 43 38. Fax +32(0)2 412 42 00. E-mail 
[log in to unmask]