R help - Dec 2013 - cumulative incidence for mstate in Survival package in R

Dear All,

I want to have the cumulative incidence curves for 'mstate' data using
Survival package in R. But I got some problems:
I. Problem 1:
1. If I only use intercept without any covariates, I can have 'right'
cumulative incidence curves (2 for 2 competing risks):
library(Survival)
fitCI <- survfit(Surv(stop, status*as.numeric(event),
type="mstate") ~
1,data=mgus1, subset=(start==0))
plot(fitCI)
2. If I include one variate ('sex'), I get 4 curves (attached; I guess
because there are two levels in 'sex' and 2 competing risks):
fitCI <- survfit(Surv(stop, status*as.numeric(event),
type="mstate")
~sex,data=mgus1, subset=(start==0))
plot(fitCI)
However, I want to just have 2 cumulative incidence curves estimated from
several covariates (such as 'sex', 'age', 'alb', etc. in
mgus1). Could you
please help me to do that? Thank you very much!
II. Problem 2:
I try using an example from sourcecode.pdf:
fit <- survfit(Surv(time, status, type=?mstate?) ~ sex, data=mine)
but where can I have the 'mine' data? Thank you!

Best,

Jieyue
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Rplot.png
Type: image/png
Size: 9373 bytes
Desc: not available
URL:
<https://stat.ethz.ch/pipermail/r-help/attachments/20131230/4a7a0a41/attachment.png>

On 12/30/2013 11:04 PM, Jieyue Li wrote:
> Dear All,
>
> I want to have the cumulative incidence curves for 'mstate' data
using
> Survival package in R. But I got some problems:
> I. Problem 1:
> 1. If I only use intercept without any covariates, I can have
'right'
> cumulative incidence curves (2 for 2 competing risks):
> library(Survival)
That shouldn't work;)
> fitCI <- survfit(Surv(stop, status*as.numeric(event),
type="mstate") ~
> 1,data=mgus1, subset=(start==0))
> plot(fitCI)
> 2. If I include one variate ('sex'), I get 4 curves (attached; I
guess
> because there are two levels in 'sex' and 2 competing risks):
> fitCI <- survfit(Surv(stop, status*as.numeric(event),
type="mstate")
> ~sex,data=mgus1, subset=(start==0))
> plot(fitCI)
> However, I want to just have 2 cumulative incidence curves estimated from
> several covariates (such as 'sex', 'age', 'alb',
etc. in mgus1). Could you
> please help me to do that? Thank you very much!
I suggest that you check the Task Views, under 'Survival' and 
'Multistate Models', for instance the 'cmprsk' and
'timereg' packages.
> II. Problem 2:
> I try using an example from sourcecode.pdf:
> fit <- survfit(Surv(time, status, type=?mstate?) ~ sex, data=mine)
> but where can I have the 'mine' data? Thank you!
Where do you find 'sourcecode.pdf'?

G?ran Brostr?m
>
> Best,
>
> Jieyue
>

Question 1: How to get just 2 cumulative incidence curves when there are
multiple covariates.
   I don't understand what you want.  Assume that we have "liver
transplant" and "death
while waiting for a transplant" as my two events.  There are overall curves
(2), or one
can create curves separately for each sex, or for different institutions.  What
do you
mean by "a curve for age"?
   If you want competing risks after Cox model adjustment, see the mstate
package.

Question 2: "mine" data.  There is no such data.  This was a
hypthetical example in the
document, and I chose a poor name for the data set; "your_data_set"
would have been
better.  I was using "mine" in the sense of "this data set is
mine, it belongs to me", and
now see that it could confuse someone.  The file sourcecode.pdf is intended to
document
the computational algorithms, but not how a user would approach the function.  A
vignette
is planned, someday...

Terry Therneau


On 12/30/2013 04:04 PM, Jieyue Li wrote:
> Dear All,
>
> I want to have the cumulative incidence curves for 'mstate' data
using Survival package in
> R. But I got some problems:
> I. Problem 1:
> 1. If I only use intercept without any covariates, I can have
'right' cumulative incidence
> curves (2 for 2 competing risks):
> library(Survival)
> fitCI <- survfit(Surv(stop, status*as.numeric(event),
type="mstate") ~ 1,data=mgus1,
> subset=(start==0))
> plot(fitCI)
> 2. If I include one variate ('sex'), I get 4 curves (attached; I
guess because there are
> two levels in 'sex' and 2 competing risks):
> fitCI <- survfit(Surv(stop, status*as.numeric(event),
type="mstate") ~sex,data=mgus1,
> subset=(start==0))
> plot(fitCI)
> However, I want to just have 2 cumulative incidence curves estimated from
several
> covariates (such as 'sex', 'age', 'alb', etc. in
mgus1). Could you please help me to do
> that? Thank you very much!
> II. Problem 2:
> I try using an example from sourcecode.pdf:
> fit <- survfit(Surv(time, status, type=?mstate?) ~ sex, data=mine)
> but where can I have the 'mine' data? Thank you!
>
> Best,
>
> Jieyue
>