Message-ID: <CAMESY2hHNw4skjcK4bo8AFxSDOr+0b=qg3R4Z3ABMeu5haVHRg@mail.gmail.com>
Date: 2018-11-21T02:34:59Z
From: Israel Ortiz
Subject: Predict follow up time using parametric model in r
In-Reply-To: <bb6fa2$adn3p3@ironport10.mayo.edu>
You are right. Specifically, I need to predict the mean and median time to
failure from a coxph model and several parametric models using new data.
Thanks.
El lun., 5 nov. 2018 a las 7:11, Therneau, Terry M., Ph.D. (<
therneau at mayo.edu>) escribi?:
> First, type='expected' gives the expected cumulative hazard for each
> subject, given their follow-up time and covariates. It is not the expected
> follow-up time, the expected time to death, or the probability of death.
> I suspect you are not getting what you think you are.
>
> A survival model predicts a survival curve for each subject. For Cox
> models you get the entire curve with the survfit() method, for survreg
> models you get the curve with predict(). To get a better answer about
> how to "predict follow-up time" you will need to be more clear about what
> you actually want, statistically. Mean time to failure? Median? RMST?
> ....
>
> Terry T.
>
>
> On 11/4/18 5:00 AM, r-help-request at r-project.org wrote:
>
> I am trying to predict follow-up time using several survival models, both
> parametric and semi-parametric. I achieve it for semi parametric models
> using predict.coxph function in R from survival package using type =
> "expected" as indicated in help. However, for parametric models, this
> option doesn't exist for the predict.survreg function. Is there any other
> option? Maybe using rms package?
>
>
>
>
[[alternative HTML version deleted]]