Software

vamiss  [2024, GitHub]
Given paired top-cause of death estimates from the VA algorithm and MITS diagnosis (gold standard), it estimates country-specific misclassification rates by adjusting for the algorithm's systematic bias and cross-country heterogeneity Pramanik et al. (2024+). It also provides estimates of intrinsic accuracy and systematic bias, offering deeper insights into the algorithm's performance and driving future refinements.
 

GPDR  [2024, GitHub]
Suppose a scalar response and multiple repeated measures are observed from individuals. Assuming the repeated measures from each individual are random samples from a covariate distribution, the software regresses the scalar response on distribution-valued covariates (Tang and Pramanik et al., 2024). It assumes a Gaussian process prior on the regression function and is invariant to any transformation or ordering of the repeated measures.
 

BFF  [2023, R package (Downloads)]
Obtains Bayes factor functions (BFFs) based on z, t, chi-squared, and F test statistics (Johnson et al., 2023). The prior densities are centered on standardized effects, which serve as indices for the BFF. They summarize evidence corresponding to a range of scientifically interesting effect sizes. BFFs are available in closed form and depend on hyperparameters, which determine the shape and scale of the prior distributions defining the alternative hypotheses.
 

NAP  [2022, R package (Downloads), GitHub]
Conducts one- and two-sample Bayesian z and t tests for a point null against a two-sided alternative using non-local alternative prior (NAP; Pramanik and Johnson, 2022). The software accommodates the normal moment prior and composite alternative as two types of NAP densities. It also computes Bayes factors and the expected weight of evidence in fixed-design tests for different effect sizes and sample sizes. Furthermore, for sequential data, the package also performs sequential tests, and calculates operating characteristics and average sample number.
 

MSPRT  [2021, R package (Downloads), GitHub]
Given the maximum available sample size (N) and the target levels of Type I and II error probabilities, the software designs MSPRT for conducting one-sample proportion tests, and one and two-sample z and t tests (Pramanik et al., 2021). It exactly maintains the Type I error and outputs the operating characteristics. The supplement provides a user's guide.