dynamichazard (version 1.0.2)

predict.ddhazard: Predict Method for ddhazard Object

Description

Predict method for ddhazard.

Usage

# S3 method for ddhazard
predict(
  object,
  new_data,
  type = c("response", "term"),
  tstart = "start",
  tstop = "stop",
  use_parallel,
  sds = FALSE,
  max_threads,
  ...
)

Value

Returns a list with elements as described in the Term and Response sections.

Arguments

object

result of a ddhazard call.

new_data

new data to base predictions on.

type

either "response" for predicted probability of an event or "term" for predicted terms in the linear predictor.

tstart

name of the start time column in new_data. It must be on the same time scale as the tstart used in the Surv(tstart, tstop, event) in the formula passed to ddhazard.

tstop

same as tstart for the stop argument.

use_parallel

not longer supported.

sds

TRUE if point wise standard deviation should be computed. Convenient if you use functions like ns and you only want one term per term in the right hand site of the formula used in ddhazard.

max_threads

not longer supported.

...

not used.

Term

The result with type = "term" is a lists of list each having length equal to nrow(new_data). The lists are

terms

It's elements are matrices where the first dimension is time and the second dimension is the terms.

sds

similar to terms for the point-wise confidence intervals using the smoothed co-variance matrices. Only added if sds = TRUE.

fixed_terms

contains the fixed (non-time-varying) effect.

varcov

similar to sds but differs by containing the whole covariance matrix for the terms. It is a 3D array where the third dimension is time. Only added if sds = TRUE.

start

numeric vector with start time for each time-varying term.

tstop

numeric vector with stop time for each time-varying term.

Response

The result with type = "response" is a list with the elements below. If tstart and tstop are matched in columns in new_data, then the probability will be for having an event between tstart and tstop conditional on no events before tstart.

fits

fitted probability of an event.

istart

numeric vector with start time for each element in fits.

istop

numeric vector with stop time for each element in fits.

Details

The function check if there are columns in new_data which names match tstart and tstop. If matched, then the bins are found which the start time to the stop time are in. If tstart and tstop are not matched then all the bins used in the estimation method will be used.

Examples

fit <- ddhazard(
 Surv(time, status == 2) ~ log(bili), pbc, id = pbc$id, max_T = 3600,
 Q_0 = diag(1, 2), Q = diag(1e-4, 2), by = 50,
 control = ddhazard_control(method = "GMA"))
predict(fit, type = "response", new_data =
 data.frame(time = 0, status = 2, bili = 3))
predict(fit, type = "term", new_data =
 data.frame(time = 0, status = 2, bili = 3))

# probability of an event between time 0 and 2000 with bili = 3
predict(fit, type = "response", new_data =
          data.frame(time = 0, status = 2, bili = 3, tstart = 0, tstop = 2000),
        tstart = "tstart", tstop = "tstop")