| Title: | Power Calculations for SNP Studies with Censored Outcomes |
|---|---|
| Description: | Conduct asymptotic and empirical power and sample size calculations for Single-Nucleotide Polymorphism (SNP) association studies with right censored time to event outcomes. |
| Authors: | Kouros Owzar, Zhiguo Li, Nancy Cox, Sin-Ho Jung and Chanhee Yi |
| Maintainer: | Alexander Sibley <[email protected]> |
| License: | GPL-3 |
| Version: | 0.26 |
| Built: | 2024-11-10 06:32:44 UTC |
| Source: | CRAN |
This package can be used to conduct asymptotic and empirical power and sample size calculations for Single-nucleotide Polymorphism (SNP) association studies with right censored time to event outcomes.
| Package: | survSNP |
| Type: | Package |
| Version: | 0.26 |
| Date: | 2023-01-20 |
| License: | GPL-3 |
| LazyLoad: | yes |
The functions sim.snp.expsurv.power and survSNP.power.table can be used to conduct power and sample size calculations. The package vignette serves as a tutorial for using this package. The technical details are provided in the reference cited below. It is highly recommended that the vignette along with this are reviewed before using this package. Currently, this package only supports additive risk models.
Kouros Owzar, Zhiguo Li, Nancy Cox, Sin-Ho Jung and Chanhee Yi
Kouros Owzar, Zhiguo Li, Nancy Cox and Sin-Ho Jung. Power and Sample Size Calculations for SNP Association Studies with Censored Time-to-Event Outcomes. https://onlinelibrary.wiley.com/doi/full/10.1002/gepi.21645
# See vignette for examples including details on the example # considered below results<-sim.snp.expsurv.power(GHR=1.25, B=0, n=500, raf=0.1, erate=0.75, pilm=0.5, lm=1, model="additive", test="additive", alpha=0.05) results[,c("n","erate","alpha","pow0")] GHRs<-seq(1.05,1.5,by=0.05) ns<-c(100,500,700) rafs<-c(0.1,0.3,0.5) erates<-c(0.5,0.7,0.9) res<-survSNP.power.table(GHRs,ns,rafs,erates,pilm=0.5,lm=1,model="additive", test="additive",alpha=0.05) # Create key for illustration KEY=paste("q=",levels(factor(res$raf)),sep="") KEY<-list(lines=list(col=1:length(KEY),lty=1:length(KEY)), text=list(labels=paste("q=",levels(factor(res$raf)),sep="")), column=3) # Illustrate Power print(xyplot(pow0~GHR|factor(erate)*factor(n),group=factor(raf), data=res,type="l",lty=KEY$lines$lty,col=KEY$lines$col, key=KEY, xlab="Genotype Hazard Ratio",ylab="Power")) # Illustrate Power (restricted to n=100) print(xyplot(pow0~GHR|factor(erate),group=factor(raf), data=subset(res,n==ns[1]), type="l",lty=KEY$lines$lty,col=KEY$lines$col, key=KEY, xlab="Genotype Hazard Ratio",ylab="Power", sub=paste("n=",ns[1],", alpha=",round(unique(res$alpha),2))))# See vignette for examples including details on the example # considered below results<-sim.snp.expsurv.power(GHR=1.25, B=0, n=500, raf=0.1, erate=0.75, pilm=0.5, lm=1, model="additive", test="additive", alpha=0.05) results[,c("n","erate","alpha","pow0")] GHRs<-seq(1.05,1.5,by=0.05) ns<-c(100,500,700) rafs<-c(0.1,0.3,0.5) erates<-c(0.5,0.7,0.9) res<-survSNP.power.table(GHRs,ns,rafs,erates,pilm=0.5,lm=1,model="additive", test="additive",alpha=0.05) # Create key for illustration KEY=paste("q=",levels(factor(res$raf)),sep="") KEY<-list(lines=list(col=1:length(KEY),lty=1:length(KEY)), text=list(labels=paste("q=",levels(factor(res$raf)),sep="")), column=3) # Illustrate Power print(xyplot(pow0~GHR|factor(erate)*factor(n),group=factor(raf), data=res,type="l",lty=KEY$lines$lty,col=KEY$lines$col, key=KEY, xlab="Genotype Hazard Ratio",ylab="Power")) # Illustrate Power (restricted to n=100) print(xyplot(pow0~GHR|factor(erate),group=factor(raf), data=subset(res,n==ns[1]), type="l",lty=KEY$lines$lty,col=KEY$lines$col, key=KEY, xlab="Genotype Hazard Ratio",ylab="Power", sub=paste("n=",ns[1],", alpha=",round(unique(res$alpha),2))))
This function calculates the asymptotic power and variance assuming that the survival distribution is a mixture of exponentials with rates and the censoring distribution is uniform on the interval (a,b).
asypow(n, theta, a, b, lambda0, q, p, alpha, z,exactvar)asypow(n, theta, a, b, lambda0, q, p, alpha, z,exactvar)
n |
Sample size |
theta |
Effect size (log genotype hazard ratio (GHR)) |
a |
Censoring distribution parameter (assumed to be uniform on [a,b]) |
b |
Censoring distribution parameter (assumed to be uniform on [a,b]) |
lambda0 |
Baseline exponential hazard rate |
q |
Relative risk allele frequency |
p |
Relative genotype frequency |
alpha |
Nominal two-sided type I error rate |
z |
Genotype scores (right now only additive scores AA=0,AB=1,BB=2 generate correct power) |
exactvar |
Indicator for using the exact variance formula |
This function is called by sim.snp.expsurv.power to calculate the asymptotic variance (exact and approximate) formulas. It is not intended to be called directly by the user. To conduct power calculations, use sim.snp.expsurv.power or the convenience wrapper function survSNP.power.table.
power |
Asymptotic power based on exact variance formula |
power0 |
Asymptotic power based on approximate variance formula |
v1 |
First term of asymptotic variance |
v2 |
Second term of asymptotic variance |
v12 |
Third term of the asymptotic variance (covariance) |
vapprox |
Approximate asymptotic variance formula (=v1) |
exact |
Exact asymptotic variance formula (=v1+v2+v12) |
diff |
Difference between variances (=v2+v12) |
ratio |
Ratio of variances (=v1/(v1+v2+v12)) |
Kouros Owzar, Zhiguo Li, Nancy Cox, Sin-Ho Jung and Chanhee Yi
Kouros Owzar, Zhiguo Li, Nancy Cox and Sin-Ho Jung. Power and Sample Size Calculations for SNP Association Studies with Censored Time-to-Event Outcomes. https://onlinelibrary.wiley.com/doi/full/10.1002/gepi.21645
This function computes the bound
censbnd(lambda, p, crate, rootint = c(0.1, 1000))censbnd(lambda, p, crate, rootint = c(0.1, 1000))
lambda |
Baseline exponential hazard rate |
p |
Relative genotype frequency |
crate |
Desired censoring rate |
rootint |
Interval to be searched for the root |
The time to event distribution is assumed to be a mixture of
exponentials with parameter ,
and with mixing proportion
, and .
Suppose that the censoring distribution is uniform on the interval
.
This function calculates for a desired censoring rate.
It is not intended to be called directly by the user.
To conduct power calculations, use sim.snp.expsurv.power
or the convenience wrapper function survSNP.power.table.
This function returns a list from the uniroot function. The
root component of this list in the
Kouros Owzar, Zhiguo Li, Nancy Cox, Sin-Ho Jung and Chanhee Yi
Kouros Owzar, Zhiguo Li, Nancy Cox and Sin-Ho Jung. Power and Sample Size Calculations for SNP Association Studies with Censored Time-to-Event Outcomes. https://onlinelibrary.wiley.com/doi/full/10.1002/gepi.21645
censbnd(0.1,hwe(0.1),0.9)$rootcensbnd(0.1,hwe(0.1),0.9)$root
Compute relative genotypic frequencies for a given relative allelic frequency
hwe(raf)hwe(raf)
raf |
Relative minor allele frequency for the B allele. |
For a bi-allelic SNP with genotypes AA, AB and
BB,
with a relative allele frequency for the B allele,
this function returns the corresponding relative genotypic
frequencies.
It is not intended to be called directly by the user.
To conduct power calculations, use sim.snp.expsurv.power
or the convenience wrapper function survSNP.power.table.
A vector of length three relative genotypic frequencies.
Kouros Owzar, Zhiguo Li, Nancy Cox, Sin-Ho Jung and Chanhee Yi
Kouros Owzar, Zhiguo Li, Nancy Cox and Sin-Ho Jung. Power and Sample Size Calculations for SNP Association Studies with Censored Time-to-Event Outcomes. https://onlinelibrary.wiley.com/doi/full/10.1002/gepi.21645
censbnd(0.1,hwe(0.1),0.9)censbnd(0.1,hwe(0.1),0.9)
This function calculates asymptotic and empirical power for SNP association studies
sim.snp.expsurv.power(GHR, B, n, raf, erate, pilm, lm, model, test, alpha, exactvar = FALSE, interval = c(0, 10), rootint = c(0.1, 200))sim.snp.expsurv.power(GHR, B, n, raf, erate, pilm, lm, model, test, alpha, exactvar = FALSE, interval = c(0, 10), rootint = c(0.1, 200))
GHR |
Genotype Hazard Ratio |
B |
Number of simulation replicates (set to 0 if no empirical calculations are desired) |
n |
Sample size |
raf |
Relative risk allele frequency |
erate |
Event Rate |
pilm |
Probability that the time-to-event is greater than lm |
lm |
Landmark time used for powering the study |
model |
True genetic risk model (choices are "additive", "recessive" or "dominant". For the asymptotic calculations only the "additive" should be used until further notice) |
test |
Hypothesized genetic risk model (choices are "additive", "recessive" or "dominant". For the asymptotic calculations only the "additive" should be used until further notice) |
alpha |
Nominal two-sided type I error rate |
exactvar |
Indicator for using the exact variance formula |
interval |
Interval to search for baseline hazard rate |
rootint |
Interval to search for censoring bound |
This function calculates asymptotic and empirical power for SNP association studies
This function returns a data.frame with the following columns
B |
Number of simulation replicates |
raf |
Relative allelic frequency of the B allele |
q0 |
Relative genotypic frequency for AA |
q1 |
Relative genotypic frequency for AB |
q2 |
Relative genotypic frequency for BB |
lam0 |
Exponential hazard rate for P(T>t|AA) |
lam1 |
Exponential hazard rate for P(T>t|AB) |
lam2 |
Exponential hazard rate for P(T>t|BB) |
GHR |
Genotype Hazard Ratio |
pilm |
Probability that the time to event in the population exceeds the landmark lm: P(T>lm) |
lm |
Landmark (see pilm above) |
model |
The true genetic risk model |
test |
The hypothesized genetic risk model |
a |
Parameter for the censoring distribution (uniform on [a,b]) |
b |
Parameter for the censoring distribution (uniform on [a,b]) |
erate |
Event rate |
n |
Sample size |
powB |
Empirical Power |
pow |
Asymptotic Power based on the exact variance formula |
pow0 |
Asymptotic Power based on the approximate variance formula |
v1 |
First term of the variance (v1+v2+v12) |
v2 |
Second term of the variance (v1+v2+v12) |
v12 |
Third term (covariance) of the variance (v1+v2+v12) |
Kouros Owzar, Zhiguo Li, Nancy Cox, Sin-Ho Jung and Chanhee Yi
Kouros Owzar, Zhiguo Li, Nancy Cox and Sin-Ho Jung. Power and Sample Size Calculations for SNP Association Studies with Censored Time-to-Event Outcomes. https://onlinelibrary.wiley.com/doi/full/10.1002/gepi.21645
# See vignette for examples including details on the example # considered below results<-sim.snp.expsurv.power(GHR=1.25, B=0, n=500, raf=0.1, erate=0.75, pilm=0.5, lm=1, model="additive", test="additive", alpha=0.05) results[,c("n","erate","alpha","pow0")]# See vignette for examples including details on the example # considered below results<-sim.snp.expsurv.power(GHR=1.25, B=0, n=500, raf=0.1, erate=0.75, pilm=0.5, lm=1, model="additive", test="additive", alpha=0.05) results[,c("n","erate","alpha","pow0")]
This function simulates the asymptotic P-value for the Cox score statistic under a specified genetic risk model
sim.snp.expsurv.sctest(n, gtprev, lam, a, b, ztest, diag = FALSE)sim.snp.expsurv.sctest(n, gtprev, lam, a, b, ztest, diag = FALSE)
n |
Sample size |
gtprev |
Relative genotypic frequency |
lam |
Exponential hazard rates for conditional time to event survival functions |
a |
Lower bound for the uniform censoring bound |
b |
Upper bound for the uniform censoring bound |
ztest |
Assumed genetic model |
diag |
Set to TRUE if print out of diagnostics is desired |
This function simulates the genotypes 0, 1 or 2 from relative
genotypic frequencies. Then conditional on genotype, it simulates the
time to event from an exponential distribution with parameter
. The censoring time is drawn from a uniform law on the
interval .
It is not intended to be called directly by the user.
To conduct power calculations, use sim.snp.expsurv.power
or the convenience wrapper function survSNP.power.table.
A vector of length 2 containing the observed event rate and the asymptotic P-value for the Cox score test
Kouros Owzar, Zhiguo Li, Nancy Cox, Sin-Ho Jung and Chanhee Yi
Kouros Owzar, Zhiguo Li, Nancy Cox and Sin-Ho Jung. Power and Sample Size Calculations for SNP Association Studies with Censored Time-to-Event Outcomes. https://onlinelibrary.wiley.com/doi/full/10.1002/gepi.21645
This function calculates the exponential hazard rates for the conditional survival functions
surv.exp.gt.model(pilm, lm, gtprev, GRR, zmodel, interval)surv.exp.gt.model(pilm, lm, gtprev, GRR, zmodel, interval)
pilm |
Probability that the time-to-event is greater than lm |
lm |
Landmark time used for powering the study |
gtprev |
Relative genotypic frequency |
GRR |
Genotype Hazard Ratio |
zmodel |
Genetic Risk Model (choices are "additive", "recessive" or "dominant") |
interval |
Interval to search for baseline hazard rate |
Determines the time to event distributions in the three component mixture model discussed in the reference below. It is not intended to be called directly by the user. To conduct power calculations, use sim.snp.expsurv.power or the convenience wrapper function survSNP.power.table.
A vector of length three containing the exponential hazard rates
Kouros Owzar, Zhiguo Li, Nancy Cox, Sin-Ho Jung and Chanhee Yi
Kouros Owzar, Zhiguo Li, Nancy Cox and Sin-Ho Jung. Power and Sample Size Calculations for SNP Association Studies with Censored Time-to-Event Outcomes. https://onlinelibrary.wiley.com/doi/full/10.1002/gepi.21645
This function produces a table with the resulting empirical and asymptotic power over a given range of Genotype Hazard Ratios, relative allelic frequencies, sample sizes, and event rates.
survSNP.power.table(GHRs, ns, rafs, erates, pilm, lm, model, test, alpha, exactvar = FALSE, B = 0)survSNP.power.table(GHRs, ns, rafs, erates, pilm, lm, model, test, alpha, exactvar = FALSE, B = 0)
GHRs |
A vector of Genotype Hazard Ratios |
ns |
A vector of sample sizes |
rafs |
A vector of relative allelic frequencies (for the risk allele) |
erates |
A vector of event rates |
pilm |
Probability that the time-to-event is greater than lm |
lm |
Landmark time used for powering the study |
model |
True genetic risk model (choices are "additive", "recessive" or "dominant". For the asymptotic calculations only the "additive" should be used until further notice) |
test |
Hypothesized genetic risk model (choices are "additive", "recessive" or "dominant". For the asymptotic calculations only the "additive" should be used until further notice) |
alpha |
Nominal two-sided type I error rate |
exactvar |
Indicator for using the exact variance formula |
B |
Number of simulation replicates (set to 0 if no empirical calculations are desired) |
This version only supports additive models.
See output of sim.snp.expsurv.power.
Kouros Owzar, Zhiguo Li, Nancy Cox, Sin-Ho Jung and Chanhee Yi
Kouros Owzar, Zhiguo Li, Nancy Cox and Sin-Ho Jung. Power and Sample Size Calculations for SNP Association Studies with Censored Time-to-Event Outcomes. https://onlinelibrary.wiley.com/doi/full/10.1002/gepi.21645
GRRs<-seq(1.5,2,by=0.25) ns<-c(100,500,1000) rafs<-c(0.3,0.5,0.7) erates=c(0.5,0.7,0.9) res<-survSNP.power.table(GRRs,ns,rafs,erates,pilm=0.5,lm=1,model="additive", test="additive",alpha=0.05) # Create key for illustration KEY=paste("q=",levels(factor(res$raf)),sep="") KEY<-list(lines=list(col=1:length(KEY),lty=1:length(KEY)), text=list(labels=paste("q=",levels(factor(res$raf)),sep="")), column=3) # Illustrate Power print(xyplot(pow0~GHR|factor(erate)*factor(n),group=factor(raf), data=res,type="l",lty=KEY$lines$lty,col=KEY$lines$col, key=KEY, xlab="Genotype Hazard Ratio",ylab="Power")) # Illustrate Power (restricted to n=100) print(xyplot(pow0~GHR|factor(erate),group=factor(raf), data=subset(res,n==ns[1]), type="l",lty=KEY$lines$lty,col=KEY$lines$col, key=KEY, xlab="Genotype Hazard Ratio",ylab="Power", sub=paste("n=",ns[1],", alpha=",round(unique(res$alpha),2))))GRRs<-seq(1.5,2,by=0.25) ns<-c(100,500,1000) rafs<-c(0.3,0.5,0.7) erates=c(0.5,0.7,0.9) res<-survSNP.power.table(GRRs,ns,rafs,erates,pilm=0.5,lm=1,model="additive", test="additive",alpha=0.05) # Create key for illustration KEY=paste("q=",levels(factor(res$raf)),sep="") KEY<-list(lines=list(col=1:length(KEY),lty=1:length(KEY)), text=list(labels=paste("q=",levels(factor(res$raf)),sep="")), column=3) # Illustrate Power print(xyplot(pow0~GHR|factor(erate)*factor(n),group=factor(raf), data=res,type="l",lty=KEY$lines$lty,col=KEY$lines$col, key=KEY, xlab="Genotype Hazard Ratio",ylab="Power")) # Illustrate Power (restricted to n=100) print(xyplot(pow0~GHR|factor(erate),group=factor(raf), data=subset(res,n==ns[1]), type="l",lty=KEY$lines$lty,col=KEY$lines$col, key=KEY, xlab="Genotype Hazard Ratio",ylab="Power", sub=paste("n=",ns[1],", alpha=",round(unique(res$alpha),2))))