Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: RE: saving coefficient matrix for two models

From   Maarten buis <>
Subject   Re: st: RE: saving coefficient matrix for two models
Date   Wed, 23 Feb 2011 08:23:35 +0000 (GMT)

--- On Wed, 23/2/11, Michael Palmer wrote:
> I'm estimating a two part model (first part logit, second
> part log linear model) and am calculating the marginal effects which
> require coefficients from both models. The question is how to save
> coefficients from the two consecutive models. 
> For example.. 
> Logit hospital_visit $xlist
> Mfx
> Mat b = e(b)
> Reg Ln_hospital_expenditure $xlist
> Mat b = e(b)
> I presume 'display _b[age]' for example will give me the
> coefficient for second model but how to derive the same
> coefficient from the first model. 

Your best chance is to use -suest-, as this will not only stack
the coefficient matrix but also creates a meaninigfull variance 
covariance matrix for that stacked coefficient matrix. This 
means that you can compute your marginal effect using -nlcom-,
which means you will also get standard errors for your marginal

Hope this helps,

Ps. the term log linear model means different things in different disciplines. In my (sub-(sub-))discipline (sociology) it refers 
to a type of ANOVA for discrete dependent variables, i.e. that
model is linear in the log odds and is related to -logit- just
as ANOVA is related to -regress-. So if you ever find yourself 
talking about this model to a sociologist and you get the 
impression that you two are not understanding each other, check
whether you are talking about the same model.

Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen


*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index