Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# st: Re: Cox regression - model comparison

 From Janet Hill To statalist@hsphsun2.harvard.edu Subject st: Re: Cox regression - model comparison Date Thu, 3 Feb 2011 14:04:57 +0000 (GMT)

```I am attempting to use Cox's regression for the situation where there are 2 main factors and a possible interaction. Several measurements are made on the same tooth and I have allowed for this with the cluster option. I wanted to see if the interaction term was significant but I cannot use lrtest with the robust option. I would be grateful for an indication of now to do this comparison. I am using Stata 11.

Many thanks.
Janet.
. stcox am##adh, nohr vce(cluster tooth)

failure _d:  fail == 1
analysis time _t:  bs

Iteration 0:   log pseudolikelihood = -2863.5777
Iteration 1:   log pseudolikelihood = -2835.7536
Iteration 2:   log pseudolikelihood = -2832.7023
Iteration 3:   log pseudolikelihood = -2832.6818
Iteration 4:   log pseudolikelihood = -2832.6818
Refining estimates:
Iteration 0:   log pseudolikelihood = -2832.6818

Cox regression -- Breslow method for ties

No. of subjects      =          540                Number of obs   =       540
No. of failures      =          540
Time at risk         =   20471.8399
Wald chi2(8)    =     33.97
Log pseudolikelihood =   -2832.6818                Prob > chi2     =    0.0000

(Std. Err. adjusted for 45 clusters in tooth)
------------------------------------------------------------------------------
|               Robust
_t |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
am |
1  |  -.6271124   .2370507    -2.65   0.008    -1.091723   -.1625016
2  |     -.3282   .2142657    -1.53   0.126     -.748153    .0917531
|
2  |   .3572293   .2726956     1.31   0.190    -.1772442    .8917028
3  |   .5961289   .2648755     2.25   0.024     .0769825    1.115275
|
1 2  |  -.0934599   .3588636    -0.26   0.795    -.7968197    .6098998
1 3  |   .0096932   .3414542     0.03   0.977    -.6595447    .6789312
2 2  |  -.2806778   .3151156    -0.89   0.373    -.8982931    .3369374
2 3  |   -.557413   .3178053    -1.75   0.079      -1.1803    .0654739
------------------------------------------------------------------------------

. estimates store a

. stcox am adh, nohr vce(cluster tooth)

failure _d:  fail == 1
analysis time _t:  bs

Iteration 0:   log pseudolikelihood = -2863.5777
Iteration 1:   log pseudolikelihood = -2843.1582
Iteration 2:   log pseudolikelihood = -2843.1528
Refining estimates:
Iteration 0:   log pseudolikelihood = -2843.1528

Cox regression -- Breslow method for ties

No. of subjects      =          540                Number of obs   =       540
No. of failures      =          540
Time at risk         =   20471.8399
Wald chi2(2)    =     22.02
Log pseudolikelihood =   -2843.1528                Prob > chi2     =    0.0000

(Std. Err. adjusted for 45 clusters in tooth)
------------------------------------------------------------------------------
|               Robust
_t |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
am |  -.3048295   .0693661    -4.39   0.000    -.4407846   -.1688745
adh |   .2128891   .0752234     2.83   0.005     .0654539    .3603243
------------------------------------------------------------------------------

. lrtest a
LR test likely invalid for models with robust vce
r(498);

*
*   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/
```