R help - Oct 2012 - Problems with coxph and survfit in a stratified model with interactions

I?m trying to set up proportional hazard model that is stratified with
respect to covariate 1 and has an interaction between covariate 1 and
another variable, covariate 2. Both variables are categorical. In the
following, I try to illustrate the two problems that I?ve encountered, using
the lung dataset.



The first problem is the warning:



To me, it seems that there are too many dummies generated.

The second problem is the error:






--
View this message in context:
http://r.789695.n4.nabble.com/Problems-with-coxph-and-survfit-in-a-stratified-model-with-interactions-tp4646095.html
Sent from the R help mailing list archive at Nabble.com.

I?m trying to set up proportional hazard model that is stratified with
respect to covariate 1 and has an interaction between covariate 1 and
another variable, covariate 2. Both variables are categorical. In the
following, I try to illustrate the two problems that I?ve encountered, using
the lung dataset.



The first problem is the warning:



To me, it seems that there are too many dummies generated.

The second problem is the error:






--
View this message in context:
http://r.789695.n4.nabble.com/Problems-with-coxph-and-survfit-in-a-stratified-model-with-interactions-tp4646096.html
Sent from the R help mailing list archive at Nabble.com.

Your df object (newdata) has to have ALL the variables in it.  Normally you
wouldn't need
the strata variable, but in this case cov1 is also a covariate.  Look at the
example I
gave earlier.

Terry T.

On 10/15/2012 05:00 AM, r-help-request at r-project.org
wrote:
> Many thanks for your very quick reply!  I tried replacing ?~ strata(cov1) +
> cov1:cov2 ? with  ?~ strata(cov1):cov2 ?. As a result, I get an error
> message saying that ?object 'cov1' not found?. I guess R refers to
the data
> frame. Any idea how to fix the error? Complete code follows.
>
> Best regards,
> Roland
>
>
>
> require(survival)
> data(lung)
> #
> lung$cov1<- as.factor(lung$ph.ecog)
> lung$cov2<- as.factor(lung$sex)
> levels(lung$cov1)[levels(lung$cov1)==0]<- "zero"
> levels(lung$cov1)[levels(lung$cov1)==1]<- "one"
> levels(lung$cov1)[levels(lung$cov1)==2]<- "two"
> levels(lung$cov1)[levels(lung$cov1)==3]<- "three"
> levels(lung$cov2)[levels(lung$cov2)==1]<- "male"
> levels(lung$cov2)[levels(lung$cov2)==2]<- "female"
> #
> df<- data.frame(
>    cov2=factor("female", levels = levels(lung$cov2))
> )
> sCox<- coxph(Surv(time, status) ~ strata(cov1):cov2, data=lung)
> sfCox<- survfit(sCox,newdata=df)
>

rm wrote
> I?m trying to set up proportional hazard model that is stratified with
> respect to covariate 1 and has an interaction between covariate 1 and
> another variable, covariate 2. Both variables are categorical. In the
> following, I try to illustrate the two problems that I?ve encountered,
> using the lung dataset.
> 
> The first problem is the warning:
> 
> To me, it seems that there are too many dummies generated.
> 
> The second problem is the error:

Adding the main effect of cov2 allows the model to run

library(survival)

lung$cov1 <- factor(lung$ph.ecog, 0:3,
c("zero","one","two","three"))
lung$cov2 <- factor(lung$sex, 1:2, c("male","female"))

sCox <- coxph(Surv(time, status) ~ strata(cov1) + cov2 + cov1:cov2,
data=lung)

df <- data.frame(
  cov1=factor(0:3, 0:3,
c("zero","one","two","three"))[1],
  cov2=factor(1:2, 1:2, c("male","female"))[1]
)
df

sfCox <- survfit(sCox, newdata=df)

plot(sfCox[1], conf=FALSE)
lines(sfCox[2], col=2)
lines(sfCox[3], col=3)
lines(sfCox[4], col=4)



--
View this message in context:
http://r.789695.n4.nabble.com/Problems-with-coxph-and-survfit-in-a-stratified-model-with-interactions-tp4646096p4646241.html
Sent from the R help mailing list archive at Nabble.com.

Dear Terry, 

Thanks for your valuable comments! Could you please restate the example that
you refer to? If I include cov1 in the data frame, as you suggest, I get the
following error message. Complete code follows.

? Error in `contrasts<-`(`*tmp*`, value = contr.funs[1 + isOF[nn]]) : 
  contrasts can be applied only to factors with 2 or more levels"

Regarding Chris?s suggestion, I would prefer not to have the main effect of
cov2 in the model. I would prefer to stick with ?~ strata(cov1):cov2? to
keep the interpretation of the regression coefficient straightforward.

Best regards,
Roland



require(survival) 
data(lung) 
# 
lung$cov1 <- as.factor(lung$ph.ecog) 
lung$cov2 <- as.factor(lung$sex) 
levels(lung$cov1)[levels(lung$cov1)==0] <- "zero" 
levels(lung$cov1)[levels(lung$cov1)==1] <- "one" 
levels(lung$cov1)[levels(lung$cov1)==2] <- "two" 
levels(lung$cov1)[levels(lung$cov1)==3] <- "three" 
levels(lung$cov2)[levels(lung$cov2)==1] <- "male" 
levels(lung$cov2)[levels(lung$cov2)==2] <- "female" 
# 
df <- data.frame( 
  cov1=factor("one", levels = levels(lung$cov1)),
  cov2=factor("female", levels = levels(lung$cov2)) 
) 
sCox <- coxph(Surv(time, status) ~ strata(cov1):cov2, data=lung) 
sfCox <- survfit(sCox,newdata=df)



--
View this message in context:
http://r.789695.n4.nabble.com/Problems-with-coxph-and-survfit-in-a-stratified-model-with-interactions-tp4646096p4646245.html
Sent from the R help mailing list archive at Nabble.com.