Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

Re: st: Predicted values where a subgroup of variables are held constant


From   Michael Harhay <[email protected]>
To   Tim Wade <[email protected]>
Subject   Re: st: Predicted values where a subgroup of variables are held constant
Date   Fri, 18 Mar 2011 10:08:39 -0400

Hi Tim,
Thank you for responding. I think this is close, but am having a tough
time translating specifically to my model. I don't think it is exactly
as I don't want to predict based on only one level of one co variate.

let me try and outline better my case and sorry if it is the same.

We are trying to predict an adjusted individual probability of being
re-hospitalized.

Our outcome is binary, but I was using OLS to try and work out the
coding for the moment.

Step 1:  I regress rehosp using patient characteristics, hospital
characteriscts and local market characteristics

   regress rehosp $patient $hospital $market

my PI would like me to do something that would work like this if the
code allowed

    predict y, xb at((means) hospital market)

So we want to keep all the controls in the global command for hospital
and market at their means but let the prediction vary at the
individual level.

-Michael




On Fri, Mar 18, 2011 at 9:53 AM, Tim Wade <[email protected]> wrote:
> Is something like this what you are looking for?
>
> sysuse auto.dta
> regress price  mpg gear_ratio length weight
> *price with covariates at means and mpg at 20
> margins, at((mean) _all mpg=20)
> *generate predicted results at means for each observation of mpg
> adjust gear_ratio length weight, gen(yhat)
> *confirm result is same for margins
> list yhat if mpg==20
>
>
> Tim
>
>
> On Thu, Mar 17, 2011 at 5:11 PM, Michael Harhay
> <[email protected]> wrote:
>> Dear all,
>> I want to do a straightforward xb prediction after an OLS regression
>> on a patient population, but for a subgroup of variables I want to use
>> the sample mean not the true individual value for the prediction:
>> I've been playing with both the "predict" and "margins" command but
>> with no success. Is there an "if" like command I can put after predict
>> to force this?
>> After no success trying that I also tried to go about this in a roundabout way:
>> preserve
>> foreach var of local varlist `pred' {
>>    sum `var’, meanonly
>>    replace `var’ = r(mean)
>>    }
>>
>> predict sev5, xb
>> The predictions are still not adjusted.
>> Any quick thoughts?
>> Thanks!
>>
>> *
>> *   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 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index