Bookmark and Share

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


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: stcox in case the ph-assumption is rejected


From   Maarten Buis <[email protected]>
To   [email protected]
Subject   Re: st: stcox in case the ph-assumption is rejected
Date   Sat, 7 Jan 2012 16:26:07 +0100

I hate to give bad news but you still haven't solved your problem: you
just replaced one model that depends on the proportional hazard
assumption with another model that relies on the proportional hazard
assumption. You still need to model how the effects changes over time.
In both -stcox- and -stpm2- that can be done by either using the
-stratify()- or the -tvc()- options. The former is mainly useful for
categorical variables that are only there as a control variable and is
not of substantive interest, the latter can be used for any
explanatory variable.

-- Maarten

On Sat, Jan 7, 2012 at 4:58 AM, Yuval Arbel <[email protected]> wrote:
> Thanks very much Alex, By scaling to hazard, it worked fine. I'm very
> satisfied with the results, which are robust to -stcox- and I
> incorporated an additional footnote in the paper draft. I'm generally
> familiar with the cubic-spline method, which allows time variations,
> but also permits well-defined derivatives at the knots. Here is the
> outputs:
>
> . doedit
>
> . do "D:\kingston\public_housing\robustness_PH_assumption.do"
>
> . clear
>
> . clear matrix
>
> . set memory 500m
> (512000k)
>
> . set matsize 800
>
> . use "g:\public housing\test_sample_May_07_Bought.dta", clear
>
> .
> .
> . stpm2 mean_reduct reductcurrent_mean_reduct rent_net8
> diff_stdmadadarea diff_mortgage permanentincomeestimate82 a
>> ppreciation,df(4) scale(hazard)
> note: delayed entry models are being fitted
>
> Iteration 0:   log likelihood = -1512.9266
> Iteration 1:   log likelihood =  -1378.564
> Iteration 2:   log likelihood = -1375.8226
> Iteration 3:   log likelihood = -1375.8201
> Iteration 4:   log likelihood = -1375.8201
>
> Log likelihood = -1375.8201                       Number of obs   =     499393
>
> ------------------------------------------------------------------------------
>             |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
> xb           |
>  mean_reduct |   .0380149   .0005223    72.79   0.000     .0369912    .0390385
> reductcurr.. |   .0209263   .0004782    43.76   0.000      .019989    .0218637
>   rent_net8 |   .0027275   .0001649    16.54   0.000     .0024043    .0030508
> diff_stdma~a |  -.2068624   .0283549    -7.30   0.000     -.262437   -.1512879
> diff_mortg~e |  -9.819042   .5364728   -18.30   0.000    -10.87051   -8.767574
> permanent~82 |  -.0005591   .0000685    -8.16   0.000    -.0006933   -.0004248
> appreciation |   23.16299   2.165592    10.70   0.000     18.91851    27.40747
>       _rcs1 |   5.573527   .1047814    53.19   0.000     5.368159    5.778895
>       _rcs2 |   1.805992   .0491048    36.78   0.000     1.709749    1.902236
>       _rcs3 |    -.34599   .0112847   -30.66   0.000    -.3681077   -.3238723
>       _rcs4 |   -.076871   .0026999   -28.47   0.000    -.0821628   -.0715792
>       _cons |   -5.34359   .1040326   -51.36   0.000     -5.54749    -5.13969
> ------------------------------------------------------------------------------
>
> . test mean_reduct==reductcurrent_mean_reduct
>
>  ( 1)  [xb]mean_reduct - [xb]reductcurrent_mean_reduct = 0
>
>           chi2(  1) =  947.55
>         Prob > chi2 =    0.0000
>
> . stcox mean_reduct reductcurrent_mean_reduct rent_net8
> diff_stdmadadarea permanentincomeestimate82 diff_mortgage a
>> ppreciation,nohr
>
>         failure _d:  fail == 1
>   analysis time _t:  time_index
>                 id:  appt
>
> Iteration 0:   log likelihood = -78368.249
> Iteration 1:   log likelihood = -74721.874
> Iteration 2:   log likelihood = -74566.501
> Iteration 3:   log likelihood = -74561.567
> Iteration 4:   log likelihood = -74561.555
> Refining estimates:
> Iteration 0:   log likelihood = -74561.555
>
> Cox regression -- Breslow method for ties
>
> No. of subjects =         9547                     Number of obs   =    499393
> No. of failures =         9547
> Time at risk    =       547035
>                                                   LR chi2(7)      =   7613.39
> Log likelihood  =   -74561.555                     Prob > chi2     =    0.0000
>
> ------------------------------------------------------------------------------
>          _t |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
> -------------+----------------------------------------------------------------
>  mean_reduct |   .0353358   .0005278    66.94   0.000     .0343012    .0363703
> reductcurr.. |   .0221957   .0005134    43.23   0.000     .0211894    .0232019
>   rent_net8 |   .0025506   .0001655    15.41   0.000     .0022263    .0028749
> diff_stdma~a |  -.4642809   .0446886   -10.39   0.000    -.5518688   -.3766929
> permanent~82 |  -.0004675   .0000689    -6.79   0.000    -.0006025   -.0003325
> diff_mortg~e |  -6.430141   .8913818    -7.21   0.000    -8.177217   -4.683064
> appreciation |   9.629971   3.161657     3.05   0.002     3.433237     15.8267
> ------------------------------------------------------------------------------
>
> . test mean_reduct==reductcurrent_mean_reduct
>
>  ( 1)  mean_reduct - reductcurrent_mean_reduct = 0
>
>           chi2(  1) =  450.86
>         Prob > chi2 =    0.0000
>
> . estat phtest,detail
>
>      Test of proportional-hazards assumption
>
>      Time:  Time
>      ----------------------------------------------------------------
>                  |       rho            chi2       df       Prob>chi2
>      ------------+---------------------------------------------------
>      mean_reduct |     -0.18814       220.31        1         0.0000
>      reductcurr..|     -0.21984       436.16        1         0.0000
>      rent_net8   |     -0.03327        10.10        1         0.0015
>      diff_stdma~a|      0.03801         0.38        1         0.5357
>      permanent~82|     -0.01174         1.35        1         0.2459
>      diff_mortg~e|      0.21517        10.31        1         0.0013
>      appreciation|     -0.05323        11.87        1         0.0006
>      ------------+---------------------------------------------------
>      global test |                    543.71        7         0.0000
>      ----------------------------------------------------------------
>
>
> On Fri, Jan 6, 2012 at 11:46 PM, Alex Gamma <[email protected]> wrote:
>> Yuval,
>>
>> provided that you -stset- your data correctly (i.e. as containing delyed entries), stpm2 obviously requires you to specifiy the scale option in order to estimate such models. Apart from the command's help-file, there is also a paper from The Stata Journal that explains the use of stpm2 in detail.
>>
>> Paul C. Lambert & Patrick Royston
>> Further development of flexible parametric models for survival analysis
>> The Stata Journal (2009) 9, Number 2, pp. 265–290
>>
>> Alex
>>
>>
>>
>>> Thanks, that sounds great.
>>>
>>> I tried this and got the following error command:
>>>
>>> . stpm2 mean_reduct reductcurrent_mean_reduct rent_net8
>>> diff_stdmadadarea diff_mortgage permanentincomeestimate82 a
>>>> ppreciation,df(4)
>>> note: delayed entry models are being fitted
>>> The scale must be specified
>>>
>>> Note that in my sample - tenants start to exercise at t=13. Is this
>>> fact has something to do with this error message?
>>>
>>> On Fri, Jan 6, 2012 at 5:14 PM, Alex Gamma <[email protected]> wrote:
>>>> Hi Yuval,
>>>>
>>>> I prefer the user-written command STPM2 for these kinds of situation. It makes it easy to model variables that violate the PH-assumption as time-dependent effects using cubic splines.
>>>>
>>>> - ssc describe stpm2 -
>>>> - ssc install stpm2 -
>>>>
>>>> Alex
>>>>
>>>>
>>>> Am 06.01.2012 um 09:06 schrieb Yuval Arbel:
>>>>
>>>>> Dear Statalist Participants,
>>>>>
>>>>> I'm working with stata 11.2. Having read carefully stata's manual
>>>>> under the title "stcox  PH-assumption tests" I have two questions
>>>>> (which seems to be relevant to Marteen's answer in another thread):
>>>>>
>>>>> The manual shows very nicely the following situation related to
>>>>> medical experiments: if we take two groups of cancer patients, where
>>>>> one group is exposed to a standard treatment and the other to a
>>>>> special treatment - and we would like to show that the experimental
>>>>> treatment is more efficient, we anticipate a paralel upward shift  of
>>>>> the projected survival rates compared to the actual ones. If this is
>>>>> the case - the PH-assumption, namely the assumption that the hazard to
>>>>> survival is constant over the sample period, is supported
>>>>> statistically.
>>>>>
>>>>> My first question is whether this discussion is relevant if I am
>>>>> applying the Cox model to describe the exercise of call (real) options
>>>>> to purchase appartments.
>>>>>
>>>>> My second question is the following: suppose that the PH-assumption
>>>>> does not hold in the sample and the above discussion is relevant. The
>>>>> stata manual says the following: "If the assumption fails, alternative
>>>>> modeling choices would be more appropriate (e.g. , a stratified Cox
>>>>> model, time-varying covariates)."
>>>>>
>>>>> The question is: is there any command to incorporate the -stcox- with
>>>>> varying hazard level across time? I'm aware of the -strata()- option,
>>>>> but I wonder whether I can somehow account for time-varying covariates
>>>>> and incorporate it with -stcox-
>>>>>
>>>>> --
>>>>> Dr. Yuval Arbel
>>>>> School of Business
>>>>> Carmel Academic Center
>>>>> 4 Shaar Palmer Street,
>>>>> Haifa 33031, Israel
>>>>> e-mail1: [email protected]
>>>>> e-mail2: [email protected]
>>>>> *
>>>>> *   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/
>>>>
>>>>
>>>> *
>>>> *   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/
>>>
>>>
>>>
>>> --
>>> Dr. Yuval Arbel
>>> School of Business
>>> Carmel Academic Center
>>> 4 Shaar Palmer Street,
>>> Haifa 33031, Israel
>>> e-mail1: [email protected]
>>> e-mail2: [email protected]
>>>
>>> *
>>> *   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/
>>
>>
>> *
>> *   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/
>
>
>
> --
> Dr. Yuval Arbel
> School of Business
> Carmel Academic Center
> 4 Shaar Palmer Street,
> Haifa 33031, Israel
> e-mail1: [email protected]
> e-mail2: [email protected]
>
> *
> *   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/



-- 
--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------

*
*   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/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index