Message-ID: <917173da-daee-4feb-a129-65a9d66dd481@k16g2000yqm.googlegroups.com>
Date: 2011-06-02T13:53:11Z
From: Mike Harwood
Subject: predict with eha package
Hello list, and thank you in advance.
I'm unable to generate predicted values when specifying newdata using
phreg and aftreg models
in the eha package, but I do not have the same problem with a
proportional hazards model from coxph. Without the newdata argument
the predicted values are returned, but with
"newdata=<model.dataframe>" coxph is fine but both aftreg and phreg
models return an "Error in predict.coxph(f.ph.eha,
newdata = mort, type = "lp") : Data is not the same size as it was in
the original fit" message. Since I ultimately want a parametric model
and the "real"
data is left truncated and right censored, I think the aftreg function
in the eha package is what I must use. Following is my sample code,
without the output.
#~ All models generated successfully ---------------------------------
f.ph <- coxph(Surv(enter, exit, event) ~ ses, data = mort)
f.ph.eha <- phreg(Surv(enter, exit, event) ~ ses, data = mort)
f.aft <- aftreg(Surv(enter, exit, event) ~ ses, data = mort)
#~ All fits generated successfully -----------------------------------
f.ph.fit <- predict(f.ph, type='lp')
f.ph.eha.fit <- predict(f.ph.eha, type='lp')
f.aft.fit <- predict(f.aft, type='lp')
#~ First fit generated successfully, others output
error--------------------
f.ph.fit <- predict(f.ph, newdata=mort, type='lp')
f.ph.eha.fit <- predict(f.ph.eha, newdata=mort, type='lp')
f.aft.fit <- predict(f.aft, newdata=mort, type='lp')
Mike