Two other possibilities.
1. Use -probpred- with the poly(#) option allows for variable of interest
to be entered in the model as a polynomial.
2. Construct a data set with the values you want before running -predict-
For example:
sysuse auto, clear
gen mpg2 = mpg^2
logit foreign price mpg mpg2 weight gear_ratio
foreach var of varlist price weight gear_ratio {
qui sum `var', meanonly
local ave_`var' = r(mean)
}
preserve
drop _all
*Predictions from 10 to 60
set obs 51
gen mpg = _n + 9
gen mpg2 = mpg^2
foreach var of newlist price weight gear_ratio {
gen `var' = `ave_`var''
}
predict pred,p
twoway line pred mpg
restore
Hope this helps,
Scott
> -----Original Message-----
> From: [email protected] [mailto:owner-
> [email protected]] On Behalf Of Nick Cox
> Sent: Tuesday, March 15, 2005 4:04 PM
> To: [email protected]
> Subject: st: RE: -prgen- with a quadratic function
>
> I don't think Scott Long is a member of Statalist.
>
> No matter: your problem should yield to -adjust-
> followed by an appropriate graph.
>
> Nick
> [email protected]
>
> Mariano Sana
>
> > Hi Statalisters. Someone experienced with ?prgen- might be
> > able to answer this.
> >
> > I ran a logit model that has, among its predictor variables,
> > AGE and AGESQ.
> > The coefficient for AGE is positive and the coefficient for AGESQ is
> > negative. The turning point is at age 79 or so. I?ve been trying to
> > generate a graph that shows how predicted probabilities vary
> > by age for a few groups in my analysis. I used the ?prgen- command from
> > the Spostado package.
> >
> > The graph that I get doesn?t make sense. The probability
> > increases nonstop
> > and becomes asymptotic to 1 at age 90 or so.
> >
> > I realize that the problem is that ?prgen- cannot vary the
> > value of AGESQ
> > as the value of AGE varies. The default for AGESQ is the
> > mean of AGESQ,
> > unless I specify another value. Obviously, I need AGESQ to
> > be the square of AGE as AGE varies.
> >
> > By now I think I have no other solution than running
> > ?prvalue- (also from
> > the Spostado package) multiple times to get the probabilities
> > I need for
> > the graph. Or is there any other way? Is this a question for
> > Scott Long?
>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
