Re: st: p value precision in clogit output

 From Luisa Foco To statalist@hsphsun2.harvard.edu Subject Re: st: p value precision in clogit output Date Thu, 13 Sep 2007 08:59:21 +0200

```Dear all,
All the best
Luisa

> oops my apologies, Sergiy is correct of course. Here is a way to do
> obtain expanded p-values for the coefficients using mata.
>
> . clogit foreign weight trunk, group(gear_ratio)
> note: multiple positive outcomes within groups encountered.
> note: 32 groups (64 obs) dropped due to all positive or
>       all negative outcomes.
>
> Iteration 0:   log likelihood = -3.4204386
> Iteration 1:   log likelihood = -2.7254521
> Iteration 2:   log likelihood = -2.7248031
> Iteration 3:   log likelihood = -2.7248029
>
> Conditional (fixed-effects) logistic regression   Number of obs   =         10
>                                                   LR chi2(2)      =       1.72
>                                                   Prob > chi2     =     0.4237
> Log likelihood = -2.7248029                       Pseudo R2       =     0.2396
>
> ------------------------------------------------------------------------------
>      foreign |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
>       weight |  -.0034601   .0033301    -1.04   0.299     -.009987    .0030668
>        trunk |    .942829   .9280568     1.02   0.310    -.8761289    2.761787
> ------------------------------------------------------------------------------
>
> . mata
> ------------------------------------------------- mata (type end to
> exit) -------------------------
> : se=sqrt(diagonal(st_matrix("e(V)")))
>
> : b=st_matrix("e(b)")
>
> : z=bt:/se
>
> : bt=b'
>
> : z=bt:/se
>
> : 2:*(1:-normal(abs(z)))
>                  1
>     +---------------+
>   1 |  .2987890939  |
>   2 |  .3096687513  |
>     +---------------+
>
>
>
>> e(p) is the value associated with the chi2 test, not the coefficients
>> of the model. The p-value that Luise quoted was associated with a
>> coefficient of the model.
>> Each coefficient will have a different p value, so the result must be
>> a vector. clogit does not save a vector of p-values. Neither do other
>> estimation commands (AFAIK).
>>
>> Best regards,
>>
>>
>>
>>
>>> Luisa,
>>>
>>> typing "ereturn list" will show all the saved results after a command.
>>> To display the p-value, you can type:
>>>
>>> di e(p)
>>>
>>> Tim
>>>
>>>
>>> On 9/12/07, Luisa Foco <luisa.foco@unipv.it> wrote:
>>>
>>>> Dear Stata users,
>>>> I have the following output, obtained with the clogit command, with a
>>>> p-value <0.0001 and I would like to display the precise value of p.
>>>>
>>>> ------------------------------------------------------------------------------
>>>>      _Icc_1 | Odds Ratio   Std. Err.      z    P>|z|     [95% Conf.
>>>> Interval]
>>>> -------------+----------------------------------------------------------------
>>>>   _112 |   1.259198   .0640371     4.53   0.000      1.13974    1.391176
>>>> ------------------------------------------------------------------------------
>>>>
>>>> Is there a general way to do this?
>>>>
>>>> Luisa
>>>>
>>>>
>>>>
>>>>
>>
>
>
>

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```