valorate (version 1.0-5)

valorate.p.value: ESTIMATES THE P-VALUE OF THE LOG-RANK TEST

Description

Estimates the p-value using specific approximations to the log-rank.

Usage

valorate.p.value.sampling(vro, vrsubo, lrv, z)
valorate.p.value.chisq(vro, vrsubo, lrv, z)
valorate.p.value.normal(vro, vrsubo, lrv, z)
valorate.p.value.gaussian(vro, vrsubo, lrv, z)
valorate.p.value.weibull(vro, vrsubo, lrv, z)
valorate.p.value.beta(vro, vrsubo, lrv, z)
valorate.p.value.all(vro, vrsubo, lrv, z)

Value

the estimated p-value (times tails).

Arguments

vro

the valorate object.

vrsubo

the subpop list object (see prepare.n1) or a numeric value representing n1 used to obtain the subpop.

lrv

if provided, the log-rank value. It is needed for .sampling, .gaussian, .weibull, .beta, .normal, and .all .

z

if provided, the log-rank value in z-score (divided by the approximated standard deviation). It is needed for .normal, .chisq, optionally to .all if normal and chisq are required.

Author

Victor Trevino

Details

This family of functions estimates the p-value of the log-rank test using specific approximations. The intended 'user' function in VALORATE is valorate.p.value.sampling, which is the function that is described in the publications. The rest of the functions are complementary for comparison with the classical approximations (chisq and normal) and for experimental purposes fitting each conditional log-rank distribution sampled (conditioned on k co-occurrences) with the specified distribution (gaussian, weibull, and beta). The function valorate.p.value.all is just a proxy to all calculations in the same function.

References

Trevino et al. 2016 https://bioinformatics.mx/index.php/bioinfo-tools/

See Also

new.valorate. valorate.survdiff. valorate.plot.empirical.

Examples

## Create a random population of 100 subjects 
## having 20 events
subjects <- numeric(100)
subjects[sample(100,20)] <- 1
vo <- new.valorate(rank=subjects, sampling.size=100000, verbose=TRUE)

groups <- numeric(100)
groups[sample(100,4)] <- 1  # only 4 subjects are within the 'mutated' group
pvr <- valorate.survdiff(vo, groups)
print(pvr)

# the same than the value of pvr
valorate.p.value.sampling(vo, vo@subpop[["subpop4"]], attributes(pvr)[[1]]["LR"])

# the same than the value of pvr
valorate.p.value.sampling(vo, 4, attributes(pvr)[[1]]["LR"])

#classical approximations:
valorate.p.value.normal(vo, 4, attributes(pvr)[[1]]["LR"], attributes(pvr)[[1]]["Z"])
valorate.p.value.chisq(vo, 4, attributes(pvr)[[1]]["LR"], attributes(pvr)[[1]]["Z"])

# approximations of the conditional log-rank sampled density
valorate.p.value.gaussian(vo, 4, attributes(pvr)[[1]]["LR"])
valorate.p.value.beta(vo, 4, attributes(pvr)[[1]]["LR"])
valorate.p.value.weibull(vo, 4, attributes(pvr)[[1]]["LR"])

# all above can be get by:
valorate.p.value.all(vo, 4, attributes(pvr)[[1]]["LR"], attributes(pvr)[[1]]["Z"])

# Estimate a p-value a given log-rank
prepare.n1(vo, 50)
valorate.p.value.all(vo, 50, 0, 0) # 0 log-rank, 0 z-score