The R package YPPE provides semiparametric modeling of lifetime data with crossing survival curves via Yang and Prentice model with piecewise exponential baseline distribution. Details about the model can be found in Demarqui and Mayrink (2019) <doi.org/10.1214/20-BJPS471>. Model fitting carried out via likelihood-based and Bayesian approaches. The package also provides point and interval estimation for the crossing survival times.
You can install the released version of YPPE from CRAN with:
install.packages("YPPE")And the development version from GitHub with:
install.packages("remotes")
remotes::install_github("fndemarqui/YPPE")This is a basic example which shows you how to solve a common problem:
library(YPPE)
#> Loading required package: survival
data(gastric)
# MLE approach:
mle <- yppe(Surv(time, status)~trt, data=gastric,
approach = "mle", init = 0, n_int = 10)
summary(mle)
#> Call:
#> yppe(formula = Surv(time, status) ~ trt, data = gastric, n_int = 10,
#> approach = "mle", init = 0)
#>
#> Short-term coefficients:
#> Estimate StdErr z.value p.value
#> trt 1.77113 0.61843 2.8639 0.004185 **
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
#> Long-term coefficients:
#> Estimate StdErr z.value p.value
#> trt -0.98230 0.29576 -3.3212 0.0008962 ***
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
#> ---
#> loglik = -582.5366 AIC = 1189.073
# Bayesian approach:
bayes <- yppe(Surv(time, status)~trt, data=gastric,
approach = "bayes", n_int = 10,
refresh = FALSE)
summary(bayes)
#> Call:
#> yppe(formula = Surv(time, status) ~ trt, data = gastric, n_int = 10,
#> approach = "bayes", refresh = FALSE)
#>
#> Short-term coefficients:
#> mean se_mean sd 2.5% 25% 50% 75% 97.5% n_eff Rhat
#> trt 1.831 0.016 0.626 0.654 1.394 1.806 2.243 3.154 1599.978 1.001
#>
#> Long-term coefficients:
#> mean se_mean sd 2.5% 25% 50% 75% 97.5% n_eff Rhat
#> trt -0.971 0.007 0.316 -1.554 -1.192 -0.983 -0.778 -0.295 2062.186 1
#>
#> ---
#> Inference for Stan model: yppe.
#> 4 chains, each with iter=2000; warmup=1000; thin=1;
#> post-warmup draws per chain=1000, total post-warmup draws=4000.