RE: st: RE: Comparing regression fits of 2 different DVs to the same IV,

[Jared Saletin <jsaletin@berkeley.edu>]

Dear Al,

Thanks for the follow up. Do you happen to know of a solution for robust regression coefficients? We've been using rreg to down-weight outliers in the dataset, so we'd like to ideally compare those coefficients.

I searched around and it seems like SUEST following RREG should do the trick, but that there's an issue in the RREG code that prevents SUEST from working:

c.f. : http://www.stata.com/statalist/archive/2012-09/msg00964.html

Thanks so very much for the help!
Jared

> Hi Jared - One way to do this is with -sureg-
> 
> . sureg (sbp hr) (dbp hr)
> 
> Seemingly unrelated regression
> ----------------------------------------------------------------------
> Equation          Obs  Parms        RMSE    "R-sq"       chi2        P
> ----------------------------------------------------------------------
> sbp               883      1    18.28341    0.0722      68.68   0.0000
> dbp               883      1    10.87193    0.0046       4.11   0.0426
> ----------------------------------------------------------------------
> 
> ------------------------------------------------------------------------------
>              |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
> sbp          |
>           hr |   .1981266   .0239078     8.29   0.000     .1512681    .2449851
>        _cons |   98.89311   2.625669    37.66   0.000     93.74689    104.0393
> -------------+----------------------------------------------------------------
> dbp          |
>           hr |   .0288202   .0142164     2.03   0.043     .0009565    .0566838
>        _cons |   69.85049   1.561311    44.74   0.000     66.79037     72.9106
> ------------------------------------------------------------------------------
> 
> . lincom [sbp]hr-[dbp]hr
> 
>  ( 1)  [sbp]hr - [dbp]hr = 0
> 
> ------------------------------------------------------------------------------
>              |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
>          (1) |   .1693064    .018291     9.26   0.000     .1334568    .2051561
> ------------------------------------------------------------------------------
> 
> 
> Al Feiveson
> 
> 
> -----Original Message-----
> From: 
> owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu
> ] On Behalf Of Jared Saletin
> Sent: Tuesday, November 27, 2012 3:47 PM
> To: 
> statalist@hsphsun2.harvard.edu
> 
> Subject: st: Comparing regression fits of 2 different DVs to the same IV
> 
> Hi all,
> 
> I was hoping to ask for a little sage stat wisdom from the list.
> 
> I have two different behavioral metrics (Y1 and Y2), and I'd like to regress each onto the same within-subject predictor (X).
> 
> I've created two models:
> 
> Y1 = B0 + B1X1;
> 
> Y2 = B0 + B1X1; 
> 
> Given that Y1 and Y2 are related but different aspects of a behavioral task, I'd like to examine whether X1 has a different predictiveness on Y1 than on Y2. I know there are tests for comparing overlapping correlation coefficients, however I'd like to be able to do it within the realm of regression. Does anyone one know if this is possible?
> 
> It seems that options would included comparing standardized regression coefficients or r-squares, but it doesn't seem like either scenario is correct in a case where the DV's differ, but the IV's are the same. 
> 
> Any help would be greatly appreciated!
> 
> Cheers,
> Jared.
> 


*
*   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/