Statalist


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

Re: st: RE: -margins- vs -adjust-


From   Philip Ender <ender97@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: RE: -margins- vs -adjust-
Date   Wed, 4 Nov 2009 10:52:23 -0800

On 4nov2009 Jay Verkuilen wrote:

>-adjust- had some nice features that -margins- seems to lack, or else maybe I just don't properly understand. For instance, -adjust- used to be >able to take a variable that wasn't in the model as an argument and generate predictions, etc., for it.
>
>Example:
>
>Run a logit model over a continuous covariate, then run -adjust- over a binned version of that covariate to make a nice table of predicted >probabilities.
>
>Let me see if I can make an example:
>
>. sysuse auto
>. gen mpg2 = 10*trunc(mpg/10)+5
>. logit foreign mpg
>. adjust, by(mpg2) pr ci vertical
>
>Which gives:
>
>Dependent variable: foreign     Command: logit
>Variable left as is: mpg
>
>mpg2          pr          lb          ub
>
>15     .152759    [.074742    .286956]
>25     .353318    [.238384    .488149]
>35     .692834    [.411283    .879263]
>45      .89765    [.545156    .984657]
>
>Key:  pr         =  Probability
>[lb , ub]  =  [95% Confidence Interval]
>
>
>In short, one of the things I liked about -adjust- was the fact that it was a little bit "hackable." :) I simply haven't had to do any kind of serious >analysis using Stata 11 recently (too much other BS to deal with, committees, etc., plus the fact that some RL things intervened) so maybe >there is a way and I'm simply not seeing it.

The following -margins- commands produce very close point estimates
but somewhat different confidence intervals:

. logit foreign mpg
. margins if mpg2==15

Predictive margins                                Number of obs   =         35
Model VCE    : OIM

Expression   : Pr(foreign), predict()

------------------------------------------------------------------------------
             |            Delta-method
             |     Margin   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       _cons |   .1576242   .0514502     3.06   0.002     .0567838    .2584647

. margins if mpg2==25

Predictive margins                                Number of obs   =         32
Model VCE    : OIM

Expression   : Pr(foreign), predict()

------------------------------------------------------------------------------
             |            Delta-method
             |     Margin   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       _cons |   .3580884    .063645     5.63   0.000     .2333465    .4828303

And so on for

. margins if mpg2==35
. margins if mpg2==45

It will be interesting to see what Stata has to say about these two approaches.

-- 
Phil Ender
Statistical Consulting Group
UCLA  Academic Technology Services

*
*   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   |   What's new   |   Site index