predictStatusProb: Probability Predictions
In ModelGood: Validation of risk prediction models

Description Usage Arguments Details Value Note Author(s) See Also Examples

Function to extract probabilistic event status predictions from various diagnostic and prognostic models with binary status response. The function has a speficic method depending on the 'class' of the object.

1 2	## S3 method for class 'glm' predictStatusProb(object,newdata,...)

`object`	A model from which predicted probabilities can be extracted for the indiviuals in newdata.
`newdata`	A data frame containing data for which the `object` can provide predict probabilities. In medical applications `newdata` will typically consist of the data of patients whose data were not used for building the model.
`...`	Additional arguments that are passed on to the current method.

The function delivers predicted probabilities tailored for the model performance measures of the package. These probabilities are extracted from a fitted model of class CLASS with the function predictStatusProb.CLASS. See help(Roc) for details.

A vector with the predicted status probability for each row in NROW(newdata).

It is rather easy to write a new predictStatusProb method, see help(Roc). However, if you do not succeed, please send me an email.

The performance, in particular when doing cross-validation where the model is evaluated many times, can be improved by supressing in the call to the model all the computations that are not needed for probability prediction, for example standard error calculations.

Thomas A. Gerds tag@biostat.ku.dk

predict,Roc

library(rms)
set.seed(7)
x <- abs(rnorm(20))
d <- data.frame(y=rbinom(20,1,x/max(x)),x=x,z=rnorm(20))
nd <- data.frame(y=rbinom(8,1,x/max(x)),x=abs(rnorm(8)),z=rnorm(8))
fit <- lrm(y~x+z,d)
predictStatusProb(fit,newdata=nd)