[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: workiging with mlogit

From   Maarten buis <>
Subject   Re: st: workiging with mlogit
Date   Mon, 4 Feb 2008 10:16:58 +0000 (GMT)

--- Peter Odigie <> wrote:
> I am working with the multinomial logit command,
> mlogit and I have these questions:
> 1.  How do I save the results (coefficients, predicted
> probabilities, marginal effects,...) from the command
> to an excel file.  I tried using "matcell" , that is
> converting the results to a matrix, so that I could
> export it to excel. But this does not work.

Each discipline, journal, and individual has his own 
prefered way of looking at the results from -mlogit-. 
So Stata can't just dump everything in excel. The 
general strategy is to create the table in Stata and 
export that table using the appropriate commands 
which differ depending how you made the table. I can 
not be more specific than this because you have not 
told us what you want to export and how you created
that table (or if you made that table at all)
> 2. I discretized the continuous dependent variables in
> order to have a categorical dependent variable.  Is
> this acceptable?

Again you need to give is more information: why do you 
do this? Generally, I would say that this is a major 
waste of information, you not only loose information 
about the distances between the values but also about 
the ordering. So generally I would so that it is not

However, there are specific cases where this could 
make sense, e.g. some think that occupations can be 
ordered along a single occupational status scale, 
while others think there just are a few distinct 
classes. If you think that you want to have classes 
but only have status, and you know how to derive 
classes from status (which is definately not trivial 
or even possible) than it might make sense to turn
the continuous variable into a categorical variable.

However, often this strategy is an imperfect fix. To
continue this example: Both status and class are 
derived from answers to the question: what is your 
job? Such question results in a job title, which is 
than recoded into either class or status (or both).
If you want class than it is much better to go back 
to the original job titles and create classes from 
those. Discretizing status into classes is something 
you would only do if the original data where lost.

Hope this helps,

Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

Support the World Aids Awareness campaign this month with Yahoo! For Good
*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index