mclements
commented
3 years ago Apologies for this late reply.
You were correct that there was a bug in the predict function. The following code should now work:
library(rstpm2)
fit = gsm(Surv(rectime,censrec==1)~hormon,data=brcancer,df=3)
plot(fit,type="af",newdata=brcancer,
exposed=function(data) transform(data,hormon=1))
pred2 <- predict(fit,type="af",newdata=brcancer, grid=TRUE,
exposed=function(data) transform(data,hormon=1),
se.fit=TRUE,full=TRUE)
with(pred2, matplot(rectime,cbind(Estimate,lower,upper),type="l",lty=c(1,2,2), col=1,
xlab="Time since treatment (days)", ylab="PAF", ylim=c(0,0.5)))
proshano
Hello,
Thank you for creating this excellent package.
I noticed something a bit odd that I was hoping you could help me with.
I am trying to calculate attributable fractions for an entire set of variables that represent harmful exposures as if they were removed from the population. The idea here is to see how much of an outcome can be prevented if these things were all corrected.
predicted.frame<-predict(flex.fit.tvc, type="af", newdata=transform(data.imputed), exposed=function(data) transform(data.imputed, var1=0, var2=0, var3=0, var4=0, var5=0), se.fit=TRUE)If I plot the point estimates of the AF from the predicted.frame object above using ggplot, I see there is a tremendous amount of variability in the point estimate:
Some of the values are 0 or slightly negative when I examine the contents of predicted.frame.
When I plot with the plot() function in the package instead of using predict(), I get this:
How would I interpret that? What do the negative values on this plot mean?
Thank you for your time - I really want to make proper use of your package.