what test to use*to analyze a categorical with continuous predictor for a categorical response?

[Trenton Lum]

Hi,

I'm a statistics newbie, so please be gentle

I'm trying to figure out what test to use to analyze a categorical concurrent with continuous predictor for a categorical response.

For example:

How is hypertension with incrementally increasing age associated with overall death?

I'm trying to determine if hypertension is associated with overall death to varying degrees depending on the patient's age
i.e. As patients get older, they are more likely to die if they have hypertension than if they did not.
I have both time to death (continuous) and event of death (categorical) data.

All and any advice will be very much appreciated! thank you!

[Carlo Lazzaro]

Trenton:
welcome to the list.
As far as I can get your query, you might be interested in something along the following lines:

. use http://www.stata-press.com/data/r14/drugtr.dta
(Patient Survival in Drug Trial)

. stcox i.drug##c.age, nolog

         failure _d:  died
   analysis time _t:  studytime

Cox regression -- Breslow method for ties

No. of subjects =           48                  Number of obs    =          48
No. of failures =           31
Time at risk    =          744
                                                LR chi2(3)       =       33.33
Log likelihood  =   -83.245435                  Prob > chi2      =      0.0000

------------------------------------------------------------------------------
          _t | Haz. Ratio   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      1.drug |     .01956   .0840494    -0.92   0.360     4.30e-06    88.91912
         age |   1.106709   .0537957     2.09   0.037     1.006138    1.217332
             |
  drug#c.age |
          1  |   1.029803   .0767888     0.39   0.694     .8897808     1.19186
------------------------------------------------------------------------------

[Trenton Lum]

Thank you much Mr. Lazzaro!^^

[Marcos Almeida]

As Carlo already pointed out, it seems the issue is fundamentally related to interaction.

Therefore, as also in the example in #2, adding an interaction term can do the trick.

Perhaps an example (from the same data set) with a graph would help as well:

. margins drug, at(age=(47(2)67))
. marginsplot