# Re: st: nestreg discrepancy

 From Eduardo Nunez To statalist@hsphsun2.harvard.edu Subject Re: st: nestreg discrepancy Date Thu, 3 Dec 2009 18:11:50 -0500

```Thank you for clarify this to me. I thought wrongly that nestreg
calculated the LR test p-value for each variable, assuming a complete
model.

Eduardo

On Thu, Dec 3, 2009 at 2:57 PM, Richard Williams
<Richard.A.Williams.5@nd.edu> wrote:
> At 01:54 PM 12/3/2009, Eduardo Nunez wrote:
>>
>> Hi everyone,
>>
>> I apologize if this is a naive question, but what I understand is that
>> the LR test p-value should be close to the p-value obtained from t (or
>> wald statistic).
>> That's why I wonder the huge discrepancy I got (the likelihood ratio
>> test p-value estimated with nestreg differs so dramatically from the t
>> statistic p-value from regress).
>> Here are the results (see block 5=fe_i and block 6=lnai....both
>> variables are continuous):
>>
>> . xi: regress fc_i edad tas_i tad_i lnhb fe_i lnai lnlinfos lnpcr
>> ingprevio i.sexo*i.acxfa hta bcrdhh diuretico_previo iecaprevia
>> bbloq_previo tnihigh
>> ...
>>     tnihigh |   6.651076   1.666292     3.99   0.000     3.381331
>>  9.920821
>>       _cons |   -1037.17   314.5672    -3.30   0.001    -1654.442
>> -419.8988
>>
>> ------------------------------------------------------------------------------
>>
>> . xi: nestreg, lr quietly: regress fc_i edad tas_i tad_i lnhb fe_i
>> lnai lnlinfos lnpcr ingprevio i.sexo*i.acxfa hta bcrdhh
>> diuretico_previo iecaprevia bbloq_previo tnihigh...
>>  |    18 | -4719.984    16.10      1   0.0001  9477.968  9571.978 |
>>  +----------------------------------------------------------------+
>
> I'm not seeing the discrepancy either.  Only the original regress command
> and the final model produced by nestreg are directly comparable.  In the
> regress command, the t value for tnihigh is 3.99.  For the final nestreg
> model, the LR chi-square for adding tnihigh is 16.10, or about 3.99^2.  It
> looks good to me.  You can't go looking at all the intermediate models and
> expect the p-values for those models (where many variables have yet to be
> entered) to be the same as the p-values in the final model, where all the
> vars have been entered.
>
>
> -------------------------------------------
> 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/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/
```