Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


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

Re: st: Interaction terms


From   josephine gakii <jogakii@yahoo.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Interaction terms
Date   Thu, 14 Apr 2011 09:15:56 -0700 (PDT)

Thanks Maarten for the suggestion. I started with the simple model probit using the example on http://www.stata.com/support/faqs/stat/mfx_interact.html. In particular, i am referring to the last example that has Foreign as the dependent variable and Turn, Dum and td as the independent variables.

I ran the command; local xb _b[turn]*`meantur' + _b[dum]*`meandum' + _b[td]*`meantur'*`meandum' + _b[_cons] (seems not to be varying across observations since it is a scalar)

This is then followed by the command; predictnl dydt=normden(`xb')* (_b[turn] + _b[td]*`meandum') in 1, se(set)

I do not get how the standard errors are being computed given that xb is a constant and does not vary.

my interest in knowing this is to enable me to manually compute the standard errors and marginal effects of my multinomial probit for each equation in the respective categories.

thanks,
Josephine


--- On Thu, 4/7/11, Maarten buis <maartenbuis@yahoo.co.uk> wrote:

> From: Maarten buis <maartenbuis@yahoo.co.uk>
> Subject: Re: st: Interaction terms
> To: statalist@hsphsun2.harvard.edu
> Date: Thursday, April 7, 2011, 2:48 AM
> --- On Thu, 7/4/11, josephine gakii
> wrote:
> > yes i used the official Stata program -mprobit-.
> Because my
> > error terms are correlated i settled on -mprobit-
> which has
> > four categories in the dependent variable . 
> 
> In that case -mprobit- is not the right method for you as
> it
> assumes that the error terms are uncorrelated.
> 
> > i concur with Marteen 
> 
> Wrong double vowel, but you are not the first to make that
> 
> mistake.
> 
> > that i have to use the delta method
> > which gives the difference in the predicted
> probabilities.
> > Unfortunately i am not conversant with the proposed
> > -asmprobit-. could anyone please explain to me the
> > -asmprobit-? 
> 
> It is hard to answer a question like that, it is just too
> open ended. Posts on statalist are not a useful medium for
> 
> giving lectures on specific methods: that is what manuals,
> articles and books are good for. Email lists like statalist
> 
> are good at answering very specific question, which could
> be: "where can I find more information about -asmprobit-?"
> 
> 
> So I am going to interpret your question like that, and
> the
> answer is that the manual entry for -asmprobit- is quite 
> detailed and contains an extensive list of references, so 
> that would be the right place to look for an answer to that
> 
> question.
> 
> > again, does this imply i cannot use the delta method
> after
> > running the -mprobit- command?
> 
> You can still use the delta method, but if you think your
> model is wrong, than the delta method is not going to
> correct
> it...
> 
> > i am also interested in understanding how the
> -predictnl- and
> > -nlcom- commands work given that when i run my
> -mprobit- 
> > (assuming it is justified to use it) i get results in
> three 
> > categories (  the dependent variable has four
> categories); 
> > since the predictnl is on each variable, how can i
> extract the
> > predictions of each variable in each of the three
> different
> > categories of my dependent variable?
> 
> The trick is that the results are stored into multiple
> equations.
> You need to use those in combination with the right
> formulas for
> the model you chose to get at the predictions you want. 
> -predictnl- is very general, but the price is that you need
> to 
> do a lot yourself. When learning how to use this I would
> start
> with a simple model (e.g. -probit-), let -predictnl-
> predict 
> stuff that the official prediction function also predicts,
> work
> on it till you get the same results as the official predict
> 
> function, than move to a more complex model, do the same
> thing,
> only than start working on getting your marginal effects
> for 
> interaction variables in the simple model, and finally do
> it
> your model of interest.
> 
> Hope this helps,
> Maarten
> 
> --------------------------
> Maarten L. Buis
> Institut fuer Soziologie
> Universitaet Tuebingen
> Wilhelmstrasse 36
> 72074 Tuebingen
> Germany
> 
> http://www.maartenbuis.nl
> --------------------------
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
> 

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index