diag_crr: Checking Fine and Gray subdistribution hazards model with...

In crskdiag: Diagnostics for Fine and Gray Model

Description

Provides with a class of analytical methods and graphical approaches for checking the assumptions of the Fine and Gray subdistribution hazards model based on the cumulative sums of residuals. It validates the model in three aspects: proportionality of hazard ratio, the linear functional form, and the link function.

Usage

diag_crr(formula, data, test = c("lin", "prop"), Nit = 20, n.sim = 1000, n.plot = 10, 
		seed = NULL, minor_included = 1)

Arguments

formula	a formula object, with the response on the left of a ~ operator, and the terms on the right. The response must be a competing risks object as returned by the Crsk function.
data	a dataset contains the time, causes of event, and the covariates. For the causes, 0 is censoring, 1 is the cause of interest, 2 is all other causes.
test	type of diagnostic method used in the test. One test at a time. "lin" generates the results of testing linear functional form for each covariate and link function. "prop" generates the results of testing proportionality for each covariate and overall proportionality.
Nit	number of iterations in model estimation. Default is 20.
n.sim	number of iterations in the analytical diagnostics. Default is 1000.
n.plot	number of plots of simulated processes used in the graphical diagnostics. Default is 10.
seed	a seed used in generating the simulated processes.
minor_included	include the minor term in FG model if TRUE. Default is TRUE.

Details

More details about the model diagnostic procedures can be found in Li, Scheike and Zhang (2015).

Value

test	type of diagnostic method used in the test.
varname	covariate names.
used	total sample size used in analysis. Subject with missing value will be excluded from analysis.
NJP	total number of unique time points for the cause of interest.
TJP	unique time points for the cause of interest leading by a zero.
n.plot	number of plots of simulated processes used in the graphical diagnostics.
beta	estimated regression coefficients.
beta.se	estimated standard errors for regression coefficients.
dlamb0	a vector of estimated cumulative baseline subdistribution hazards at observed failure times.
pval	p-values of the model diagnostic test.
mav	maximum value of the cumulative residual process (lin) or standardized residual process (prop).
B	cumulative residual process (lin) or standardized cumulative residual process (prop). Used for plot functions.
uniX	unique values of the covariates. Used for plot and available only when test="lin".
TC	number of unique covariates. Available only when test="lin".

Author(s)

Jianing Li

References

Li, Jianing, Thomas H. Scheike, and Mei-Jie Zhang. "Checking Fine and Gray subdistribution hazards model with cumulative sums of residuals." Lifetime data analysis 21.2 (2015): 197-217.

Fine, Jason P., and Robert J. Gray. "A proportional hazards model for the subdistribution of a competing risk." Journal of the American statistical association 94.446 (1999): 496-509.

See Also

diag_lin, diag_prop, plot.diaglin, plot.diagprop, print.diaglin, print.diagprop

Examples

data(dat1)
out1 <- diag_crr(Crsk(time,cause)~z1+z2,data=dat1,test="lin",seed=1234)
print(out1)
plot(out1)

data(dat2)
out2 <- diag_crr(Crsk(time,cause)~z1+z2,data=dat2,test="prop",seed=1234)
print(out2)
plot(out2)

Example output

Loading required package: cmprsk
Loading required package: survival

***P-values for model diagnostics***

Linear functional form for a single covariate:
   z1    z2 
0.177 0.007 

Link function:
[1] 0.297

***P-values for model diagnostics***

Proportionality for a single covariate:
   z1    z2 
0.358 0.001 

Overall proportionality:
[1] 0.044

crskdiag documentation built on May 1, 2019, 10:53 p.m.