R help - Aug 2011 - Rpart modelling a decisión tree and getting probability

Hello everyone,

I working in a public health project and we have created a Decision Tree for
categorical variables usign the package rpart. Our goal is to develop a model
(Using the ROC tool) in order to predict presence/ausent of  diabetes and get a
better understanding  of what are the important factors in a particular chilean
population.  There are some importants variable that we have found.  Now we want
to apply this model over a big dataset in order to determinate a possible
outcome (probability of getting the deseasse), but we only have the combination
of predictive variables for a particular person.

We have created this code:

library( rpart)
fit1 <- rpart(sickness~ aetinghabit+gse+age+sex,   method="class",
data=data)
prediccion<-predict(fit1,bigdatabase, type="prob")  
predictionsyes<-prediccion[,2]
pred <- prediction(predictionsyes, datos$sickness) # but this is 



My question is. How do I put the people's conditions in this model in order
to get the people probability of getting this desease? It's possible to do a
ROC curve using only this bigdatabase? Because we don't have the outcome if
this people got or not this disease.

It would be very helpful if someone can give us some light about it. Any web
source of doing it will be very appreciated.

Thanks in advance.
Best Regards,

José Bustos
Escuela de Enfermeria
Pontificia Universidad Católica de Chile
Proyecto FONIS 2010
Celular 95939144






	[[alternative HTML version deleted]]