Re: st: Odd ratio / relative risk in logistic regression

[David Hoaglin <dchoaglin@gmail.com>]

Hi, Wong.

When you screen variables in Step 1, p < .25 is a better threshold for
including a variable in the logistic regression model.  "Any variable
whose univariate test has a p-value < 0.25 is a candidate for the
multivariable model along with all variables of known clinical
importance." (Hosmer and Lemeshow 2000, p. 95)  This strategy allows
combinations of variables to make a significant contribution to the
multivariable model when their individual relations to the outcome do
not achieve significance.

I am reluctant to comment on your output without seeing the exact
command that produced it.  I have used -logit- (which reports
coefficients) and -logistic- (which reports odds ratios), but not
-binreg-.

David Hoaglin

Hosmer DW, Lemeshow S (2000).  Applied Logistic Regression, 2nd ed.
John Wiley & Sons.

On Tue, Apr 9, 2013 at 12:06 AM, Ching Wong
<ching.y.wong@student.adelaide.edu.au> wrote:
> Hi,
>
> My analysis involves two steps:
>
> 1. Chi-square testing:
> I did a few chi-sqare testing with different variables.
> -tab grade var1, chi2
> -tab grade var2, chi2
> -tab grade var 3, ch2 etc.
> Basesd on the result of the chi-sqaure testings, the variables which
> are significant (i.e. p<0.05) will then put into the logistic
> regression.
>
> 2. logistic regression:
> I put the command as followings:
> - binreg grade var1 var3 var4 etc.
> And I have got the following output.
>
> Iteration 1:   deviance =  113.0721
> Iteration 2:   deviance =  92.10798
> Iteration 3:   deviance =  87.45499
> Iteration 4:   deviance =  86.88055
> Iteration 5:   deviance =  86.86395
> Iteration 6:   deviance =  86.86393
> Iteration 7:   deviance =  86.86393
> Generalized linear models                          No. of obs      =       297
> Optimization     : MQL Fisher scoring              Residual df     =       294
>                    (IRLS EIM)                      Scale parameter =         1
> Deviance         =  86.86392755                    (1/df) Deviance =  .2954555
> Pearson          =  311.8670508                    (1/df) Pearson  =  1.060772
> Variance function: V(u) = u*(1-u/1)                [Binomial]
> Link function    : g(u) = ln(u/(1-u))              [Logit]
>                                                    BIC             = -1587.093
> ------------------------------------------------------------------------------
>              |                 EIM
> grade |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
> var1 |   2.955512   1.066853     2.77   0.006     .8645186    5.046506
>  var4|   .4058033    1.07797     0.38   0.707     -1.70698    2.518587
>        _cons |  -4.464928   .6125685    -7.29   0.000    -5.665541   -3.264316
> ------------------------------------------------------------------------------
>
>
> In this case, I can tell var 1 is significant in the logistic
> regression model, since it has a p-value =0.006. However, how can I
> find out the odd ratio or the relative risk of this model? Did I use
> the wrong command?
*
*   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/