PTE provides bootstrap inference for evaluating personalized treatment recommendation models. It estimates how much better a proposed personalized policy performs than business-as-usual treatment assignment and quantifies the uncertainty in that improvement.
PTE supports continuous, incidence, and survival outcomes, custom performance metrics, parallel bootstrap computation, and data-driven selection of the m-out-of-n bootstrap size.
Installation
Install the current release from the Kapelner R-universe:
install.packages(
"PTE",
repos = c(
"https://kapelner.r-universe.dev",
"https://cloud.r-project.org"
)
)Alternatively, install the development version directly from GitHub:
# install.packages("pak")
pak::pak("kapelner/PTE/PTE")Example
library(PTE)
data(continuous_example)
results <- PTE_bootstrap_inference(
continuous_example$X,
continuous_example$y,
regression_type = "continuous",
B = 250,
num_cores = 1
)
summary(results)For production inference, use a larger number of bootstrap replicates, such as the function’s default. See select_optimal_m_prop() to tune the m-out-of-n bootstrap size for a particular dataset.
Citation
If PTE contributes to published work, please cite:
Kapelner, A., Bleich, J., Levine, A., Cohen, Z. D., DeRubeis, R. J., and Berk, R. (2021). Evaluating the Effectiveness of Personalized Medicine With Software. Frontiers in Big Data, 4, 572532. https://doi.org/10.3389/fdata.2021.572532
Both the methodology article and a version-specific software citation are available from R with citation("PTE"). PTE’s persistent CRAN package DOI is https://doi.org/10.32614/CRAN.package.PTE.