RE: st: How to use the Wald test when testing for interactions and heterogeneity

[Amal Khanolkar <Amal.Khanolkar@ki.se>]

Sorry!

I think the error msg crops up because I didn't specify that one of the variables is continuous! It works fine now  :)

 regress syst i.ethnicity_bi2x bwtgestage_sd i.ethnicity_bi2x#c.bwtgestage_sd age byear i.conscript_office if bmi!=. & height_cons!=. & education!=. & multibirth==1 & sex=
> =1, cluster (mor_lopnr)


Linear regression                                      Number of obs =  406833
                                                       F( 10,341530) = 1497.67
                                                       Prob > F      =  0.0000
                                                       R-squared     =  0.0353
                                                       Root MSE      =  10.869


                                           (Std. Err. adjusted for 341531 clusters in mor_lopnr)
------------------------------------------------------------------------------------------------
                               |               Robust
                          syst |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------------------------+----------------------------------------------------------------
                ethnicity_bi2x |
                            2  |  -.1666134    .112247    -1.48   0.138    -.3866144    .0533875
                            3  |  -3.942219   .2283697   -17.26   0.000    -4.389816   -3.494621
                               |
                 bwtgestage_sd |  -.1459854   .0177271    -8.24   0.000      -.18073   -.1112408
                               |
ethnicity_bi2x#c.bwtgestage_sd |
                            2  |   .3141344    .114124     2.75   0.006     .0904547    .5378141
                            3  |   .4474396   .2303636     1.94   0.052    -.0040664    .8989457
                               |
                           age |   .4489594   .0384143    11.69   0.000     .3736685    .5242502
                         byear |   .1860315   .0040078    46.42   0.000     .1781763    .1938868
                               |
              conscript_office |
                            2  |   -3.07841   .0497884   -61.83   0.000    -3.175994   -2.980826
                            3  |   1.991954   .0537103    37.09   0.000     1.886683    2.097224
                            5  |  -2.619412   .0475503   -55.09   0.000    -2.712609   -2.526214
                               |
                         _cons |  -245.8701   7.978256   -30.82   0.000    -261.5073    -230.233
------------------------------------------------------------------------------------------------


. testparm i.ethnicity_bi2x#c.bwtgestage_sd

 ( 1)  2.ethnicity_bi2x#c.bwtgestage_sd = 0
 ( 2)  3.ethnicity_bi2x#c.bwtgestage_sd = 0


       F(  2,341530) =    5.61
            Prob > F =    0.0037


Do you know how to interpret the zeros above? The P value above (0.0037) is for the overall interaction right?


Thanks,

Amal 


At 07:59 AM 9/18/2013, Amal Khanolkar wrote:
>Hi,
>
>Could anyone tell me the right syntax for using the Wald test for
>the following regression models?  I tried looking at the help page
>for the command 'test' but I wasn't able to figure it out:
>
>
>Ideally I would test for heterogeneity for the following model as
>follows using the LRT, but this doesn't go with the cluster option:
>
>regress syst i.ethnicity_bi2x age byear i.conscript_office if bmi!=.
>& height_cons!=. & education2!=. & bwtgestage_sd!=. &
>gestwk_catx2!=. & multibirth==1 & sex==1, cluster (mor_lopnr)
>estimates store a
>regress syst ethnicity_bi2x age byear i.conscript_office if bmi!=. &
>height_cons!=. & education2!=. & bwtgestage_sd!=. & gestwk_catx2!=.
>& multibirth==1 & sex==1, cluster (mor_lopnr)
>lrtest a

So, are you basically trying to test whether ethnicity_bi2x can be
treated as a continuous variable? Does something like this do what you want?

sysuse auto, clear
reg price mpg c.rep78 i.rep78
testparm i.rep78

If the test comes up insignificant (as it does in this case) this
means that the continuous version alone is sufficient.

>And similarly when testing for interactions:
>
>xi: regress syst i.ethnicity_bi2x*bwtgestage_sd age byear
>i.conscript_office bmi height_cons i.education if multibirth==1 &
>sex==1, cluster (mor_lopnr)
>estimates store A
>xi: regress syst i.ethnicity_bi2x bwtgestage_sd age byear
>i.conscript_office bmi height_cons i.education if multibirth==1 &
>sex==1, cluster (mor_lopnr)
>lrtest A

Why the switch to xi? If you have Stata 11 or higher, I think it
should be something like

regress syst i.ethnicity_bi2x bwtgestage_sd
i.ethnicity_bi2x#c.bwtgestage_sd age byear i.conscript_office bmi
height_cons i.education if multibirth==1 & sex==1, cluster (mor_lopnr)

testparm i.ethnicity_bi2x#bwtgestage_sd

You may also wish to see this thread from 2 years ago:

http://www.stata.com/statalist/archive/2011-10/msg01350.html

http://www.stata.com/statalist/archive/2011-10/msg01352.html

http://www.stata.com/statalist/archive/2011-10/msg01389.html

-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW:    http://www.nd.edu/~rwilliam

*
*   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/