[Mediation-information] question about R mediation package

[Kentaro Hirose]

Hi Yongyue,

1) If your mediator is survival time, then you should use the variable for
survival time, not the one for censorship.

2) Again, if your outcome is survival time, then you should use the
variable for survival time, not the one for censorship.


Best,
Kentaro

2016-03-13 0:56 GMT+09:00 Kosuke Imai <kimai at princeton.edu>:

> I'm ccing my colleague, Kentaro, who should be able to answer this
> question.
>
> Kosuke Imai
> Professor, Department of Politics
> Center for Statistics and Machine Learning
> Princeton University
> http://imai.princeton.edu
>
> On Wed, Mar 9, 2016 at 8:16 PM, Wei, Yongyue <ywei at hsph.harvard.edu>
> wrote:
>
>> Hi Kosuke,
>>
>>
>>
>> Sorry to trouble you! I am a user of R mediation package. Now I have one
>> question about how to use under the following situation of two survreg
>> models.
>>
>>
>>
>> model.m = survreg(Surv(pfs, progress)~group, data=ds)
>>
>> model.y = survreg(Surv(os, death)~group+progress, data=ds)
>>
>>
>>
>> med = mediate(model.m, model.y, treat="group", mediator="progress",
>>
>>         outcome="death", bootstrap=TRUE, sims = 1000)
>>
>>
>>
>>
>>
>> My question is:
>>
>> 1)      In the outcome model (model.y), which name should I use for
>> mediator? ‘progress’? or ‘pfs’?
>>
>> 2)      In mediate model, which should I use for outcome? Should be
>> ‘death’, right?
>>
>>
>>
>>
>>
>> Thank you!
>>
>>
>>
>>
>>
>>
>>
>> Best,
>>
>>
>>
>>
>>
>> Yongyue
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/mediation-information/attachments/20160314/bab3ed6f/attachment.html>