Cprob (version 1.4.1)

pseudocpf: Pseudo values for the conditional probability function

Description

The function computes pseudo values and then fit a proportional-odds model to the conditional probability function using GEE

Usage

pseudocpf(formula, data, id, subset, na.action, timepoints,
          failcode = 1, ...)

Arguments

formula

A formula object, whose terms are on the right of a ~ operator and the response, a Hist object, on the left

data

A data frame in which to interpret the formula

id

Individual patient id

subset

Expression specifying that only a subset of the data set should be used

na.action

A missing data filter funtion applied to the model.frame, after any subset argument has been used. Default is options()$na.action

timepoints

Time points at which to compute the pseudo values

failcode

Integer that specifies which event is of interest

…

Other arguments for the geese function

Value

Returns an object of class pseudocpf containing the following components:

fit

A geese object

pseudo

The pseudo values computed at the specified time points

timepoints

Same as in the function call

call

The matched call

Details

The regression model is fitted using a method based on the pseudo-values from a jackknife statistic constructed from the conditional probability curve. Then a GEE model is used on the pseudovalues to obtain the odds-ratios.

References

P.K. Andersen, J.P. Klein and S. Rosthoj (2003). Generalised Linear Models for Correlated Pseudo-Observations, with Applications to Multi-State Models. Biometrika, 90, 15-27.

J.P. Klein and P.K. Andersen (2005). Regression Modeling of Competing Risks Data Based on Pseudovalues of the Cumulative Incidence Function. Biometrics, 61, 223-229.

See Also

geese, summary.pseudocpf

Examples

# NOT RUN {
data(mgus)

cutoffs <- quantile(mgus$time, probs = seq(0, 1, 0.05))[-1]

### with fancy variance estimation
fit1 <- pseudocpf(Hist(time, ev) ~ age + creat, mgus, id = id,
                  timepoints = cutoffs, corstr = "independence",
                  scale.value = TRUE)
summary(fit1)

### with jackknife variance estimation
fit2 <- pseudocpf(Hist(time, ev) ~ age + creat, mgus, id = id,
                  timepoints = cutoffs, corstr = "independence",
                  scale.value = TRUE, jack = TRUE)
summary(fit2)
# }