Hello everyone

I would like to ask a query about 'best subsets' in Cox regression. 

When using best subsets variable selection in *multiple regression*  if
we compare models of different sizes we can use the statistics
R-sq(adj), MSE and Cp to determine a 'suitable' model in the statistical
sense.

For *survival modelling* using Cox regression there is an option for
best subsets in SAS. 

Say we have 12 candidate predictors and we request that three models of
each size should be printed,  the package will generate 3 models with 1
predictor, 3 models with 2 predictors, 3 models with 3 predictors etc.
The 'best' model is based on the global chi squared statistic.  If two
models have the same number of variables, the best model is the one with
the largest global chi squared statistic.

Now does anyone know how do we choose the 'best' model overall?  i.e.
which would be the best model if we have the following output (below)?
The manual is a little vague.

Many thanks,
Kim.

Regression Models Selected by Score Criterion

Number of       Score
Variables  Chi-Square  Variables Included in Model

        1      6.4174  pnod
        1      5.3917  nodord
        1      4.9045  dage40

        2     11.2277  difng33 pnod
        2     11.0591  pnod dage40
        2     10.4046  difng33 nodord

        3     15.4714  pnod dage40 cmvpos
        3     15.3138  difng33 pnod dage40
        3     14.9648  pnod dage40 ftom

        4     19.3362  difng33 pnod dage40 cmvpos
        4     18.8174  pnod dage40 ftom cmvpos
        4     18.6466  difng33 pnod dage40 ftom

        5     22.1963  difng33 pnod dage40 ftom cmvpos
        5     21.8328  difng33 pnod dage40 ftom tcell
        5     21.2103  difng33 nodord dage40 ftom tcell

        6     24.1018  difng33 pnod dage40 ftom cmvpos tcell
        6     23.6548  difng33 dnod nodord dage40 ftom tcell
        6     23.3141  difng33 nodord dage40 ftom cmvpos tcell

        7     25.4732  difng33 dnod nodord dage40 ftom cmvpos tcell
        7     25.2016  difng33 pnod nodord dage40 ftom cmvpos tcell
        7     24.6031  dvdr11 difng33 pnod dage40 ftom cmvpos tcell

        8     26.2360  dvdr11 difng33 dnod nodord dage40 ftom cmvpos
tcell
        8     26.1088  difng33 pnod dnod nodord dage40 ftom cmvpos tcell
        8     25.7520  dvdr11 difng33 pnod nodord dage40 ftom cmvpos
tcell

        9     26.8345  dvdr11 difng33 pnod dnod nodord dage40 ftom
cmvpos tcell
        9     26.4488  dvdr11 dil6gg difng33 dnod nodord dage40 ftom
cmvpos tcell
        9     26.2768  dvdr11 difng33 pat26c dnod nodord dage40 ftom
cmvpos tcell

       10     27.0339  dvdr11 dil6gg difng33 pnod dnod nodord dage40
ftom cmvpos tcell
       10     26.9042  dvdr11 difng33 pat26c pnod dnod nodord dage40
ftom cmvpos tcell
       10     26.8377  dvdr11 difng33 pnod dnod nodord dage40 page35
ftom cmvpos tcell

       11     27.1351  dvdr11 dil6gg difng33 pat26c pnod dnod nodord
dage40 ftom cmvpos tcell
       11     27.0376  dvdr11 dil6gg difng33 pnod dnod nodord dage40
page35 ftom cmvpos tcell
       11     26.9081  dvdr11 difng33 pat26c pnod dnod nodord dage40
page35 ftom cmvpos tcell

       12     27.1399  dvdr11 dil6gg difng33 pat26c pnod dnod nodord
dage40 page35 ftom cmvpos tcell