RE: st: prvalue after gologit2

[Francesco Acciai <francesco_takiki@hotmail.it>]

Thank you for your help.
I will probably end up specifying the weighted means for the rest of my variables, even if this was my last option, as I will have to specify 24 variables across 11 countries (in my previous email I simplified my model in order to make things easier).
 
The svy option works well and as expected gives the same results as the command [pw=weights], but I don't know how to use it for this specific purpose, or how to combine it with "margins" in order to hold all variables to weighted means with the command "prvalue", as someone suggested in a previous Statalist inquiry (object "holding variables to weighted means with prvalue"). 



> From: richardwilliams.ndu@gmail.com
> Subject: Re: st: prvalue after gologit2
> Date: Sat, 20 Jul 2013 23:51:21 +0200
> To: statalist@hsphsun2.harvard.edu
> 
> I am out of the country and can't check everything out, but a few thoughts:
> 
> gologit2 does not support the svy: prefix command but it does support an svy option. See the help.
> 
> You could just explicitly specify the values for the rest of your values, e.g. Compute the weighted means first and then set x3 and x4 to those values.
> 
> You can probably use the margins command. It is a little tedious because gologit2 does not support factor variables and (just like ologit and mlogit) you have to write a separate command for each outcome.
> 
> Sent from my iPad
> 
> On Jul 20, 2013, at 6:25 PM, Francesco Acciai <francesco_takiki@hotmail.it> wrote:
> 
>> After I run a gologit2 model (using sampling weights) I use the command prvalue to get the expected probabilities.
>> However, the system holds all variables at their unweighted means, while I would like to use the weighted means. 
>> 
>> This is my syntax:
>> 
>> xi: gologit2 Y i.x1 i.x2 x3 x4 [pw=WEIGHTS], autofit(0.001) 
>> 
>> prvalue, x(x1_2=0, x1_3=0) rest(mean) 
>> prvalue, x(x1_2=1, x1_3=0) rest(mean) 
>> prvalue, x(x1_2=0, x1_3=1) rest(mean) 
>> 
>> x1 is a variable indicating the region of residence. I would like to get the expected probabilities for the same person (an hypothetical person with a fixed set of characteristics, corresponding to the "average" person in my sample), across these three regions of residence. 
>> 
>> I saw another inquiry about holding variables to weighted means with prvalue but here prvalue is run after gologit2, which doesn't allow the use of the command "svy". 
>> 
>> So, my question is: how can I set all the variables at their weighted means when using "prvalue" after "gologit2"?
>> 
>> Thank you very much,
>> Francesco. 
>> *
>> * For searches and help try:
>> * http://www.stata.com/help.cgi?search
>> * http://www.stata.com/support/faqs/resources/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/faqs/resources/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/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/