Pseudo values for the conditional probability function

Share:

Description

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

Usage

1
2
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

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.

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

Note

Besides the estimated regression coefficients, the function returns the computed pseudo-values, so that one can fit a different model, e.g., with a different link function.

Author(s)

Arthur Allignol, arthur.allignol@uni-ulm.de

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

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
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)