# Re: st: multilevel ordered logit cond prob in gllapred

 From "INAGAMI,SANAE " To statalist@hsphsun2.harvard.edu Subject Re: st: multilevel ordered logit cond prob in gllapred Date Thu, 24 Mar 2005 09:30:00 -0800

```thanks. I'll try and see how it goes.  thanks so much.
S
Quoting Michael Ingre <Michael.Ingre@ipm.ki.se>:

> I did something similar a while ago using -nlcom- on an estimated
> ordered ologit with -gllamm-. An example of how I did it is presented
> below using auto.dta.
>
> // Michael
>
> clear
> sysuse auto
>
> // Create level 2 ID
> gen make2 = substr(make, 1, index(make, " "))
> egen pick =tag(make2)
> gen id=_n if pick
> sort make2 id
> replace id = id[_n-1] if id == .
>
> // Make turn>40 a dichotomous predictor variable
> gen turn40 = turn > 40
>
> // estimate gllamm model on rep78 with turn40 as predictor
>
> // Use local macros to temporary hold your input values
> local pv1 = 1  // predictor value 1
> local pv2 = 0  // predictor value 2
> local cut = 11 // Choose cut point i.e. which level of the dependent
> variable you want to predict
>
> // calculate log of the ratio of the two probabilities predicted from
> pv1/pv2
> nlcom ln(     (exp(`pv1' * [rep78]turn40 - [_cut`cut']_cons) /      ///
>            (1 + exp(`pv1' * [rep78]turn40 - [_cut`cut']_cons)))  /
> ///
> 	         (exp(`pv2' * [rep78]turn40 - [_cut`cut']_cons) /      ///
> 	     (1 + exp(`pv2' * [rep78]turn40 - [_cut`cut']_cons))))
>
> // Store the results from -nlcom-
> matrix b = r(b)
> matrix V = r(V)
>
> // Values needs to be exponentiated to describe RRR
> di "RRR: " exp(b[1,1])
> di "Lower 95% CI " exp(b[1,1] - sqrt(V[1,1])*1.96)
> di "Upper 95% CI " exp(b[1,1] + sqrt(V[1,1])*1.96)
>
>
>
> On 2005-03-23, at 20:49, INAGAMI,SANAE wrote:
>
> > I have looked at the gllamm manual.  I just don't know how to write
> the
> > commands.
> >
> > I want to know the probability of each ordered outcome conditional on
> a
> > categorical predictor variable being 1.
> >
> > Then with the same command, I want to change the predictor variable
> to
> > 0.
> >
> > Then I would like to set up a relative risk of outcomes by creating a
> > ratio of the probabilities (1 vs 0).
> >
> > Then I would like to come up with confidence intervals for those
> > ratios.
> >
> > I can do this in ordered logistic stata, but I can't do it with glamm
> > because I think I set up the syntax incorrectly.
> >
> > Thanks.
> > Sanae
> > *
> > *   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/
> >
> >
> ------------------------------------------------
> Michael Ingre , PhD student & Research Associate
> Department of Psychology, Stockholm University &
> National Institute for Psychosocial Medicine IPM
> Box 230, 171 77 Stockholm, Sweden
>
> *
> *   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/
```