PC.GLM: Fit a partly conditional GLM (logistic) model
In mdbrown/partlyconditional: Partly Conditional Logistic and Cox models

Description Usage Arguments Value See Also Examples

Fit a partly conditional (PC) logistic model. PC models are helpful predictive tools for (medical) contexts where long-term follow-up is available and interest lies in predicting patients’ risks for a future adverse outcome using repeatedly measured predictors over time. These methods model the risk of an adverse event conditional on survival up to a landmark time and information accrued by that time.

1	PC.GLM(id, stime, status, measurement.time, predictors, data, prediction.time)

`id`	name of numeric subject id in data
`stime`	name of survival time, may be repeated across subj. id.
`status`	name of survival status indicator, generally 1 = bad outcome/death, 0 alive/censored.
`measurement.time`	name of time of measurement from baseline.
`data`	data.frame with id, stime, status, measurement time and marker variables. Observations with missing data will be removed.
`prediction.time`	numeric value for the prediction time of interest to fit the PC logistic model.
`markers`	character vector consisting of marker names to include.

An object of class "PC_GLM" which is a list containing:

`model.fit`	A 'glm' object . Please note that the estimates of standard error associated with the model coefficients DO NOT incorporate the variation due to marker smoothing using BLUPs.
`variable.names`	vector of variable names used to fit the model.
`call`	Function call.

#'#'

prediction.time

Inputs from function call.

BLUP PC.Cox

data(pc_data)
#'#log transform measurement time for use in models.
pc_data$log.meas.time <- log10(pc_data$meas.time + 1)

pc.glm.1 <-  PC.GLM(
 id = "sub.id",
 stime = "time",
 status = "status",
 measurement.time = "meas.time",
 predictors = c("log.meas.time", "marker_1", "marker_2"),
 prediction.time = 24,
 data = pc_data)

pc.glm.1

pc.glm.1$model.fit #direct access to the glm model object

#see function BLUP to fit mixed effects models and obtain BLUP-smoothed predictors.