Software

• Randomized Multiarm Bandits (rMAB) – R code for improved adaptive data collection in multi-arm bandits.
• Rgpt: R Interface to ChatGPT – R package providing an API for interacting with ChatGPT from R.
  Show installation & example

  Setup: obtain an OPENAI_API_KEY from OpenAI and put it in .Renviron as OPENAI_API_KEY=<your key>.

  library(devtools)
  install_github("zhaozhg81/Rgpt")

  library(Rgpt)
  Rgpt("Using R, write a sample code for linear model.")
  x = matrix(rnorm(12), nrow = 3, ncol = 4)
  LatexTable(3, 4, t(round(x, digits = 3)))

• gSNR: Generalized Signal-to-Noise Ratio – R package for generalized signal-to-noise ratio and associated tests.
  Show installation & example

  library(devtools)
  install_github("zhaozhg81/gSNR")

  library(gSNR)
  n     <- 1000
  H     <- 20
  m     <- n / H
  sigma <- 1
  
  p     <- 20
  
  set.seed(2)
  D     <- 1
  beta  <- array(0, c(p, D))
  beta  <- rnorm(p, 0, 1)
  
  X <- matrix(rnorm(p * n), c(n, p))
  
  Z <- X %*% beta
  Y <- sin(Z[, 1]) * exp(Z[, 1]) * 0.2 + sigma * rnorm(n, 0, 1)
  
  res    <- gSNR(X, Y, discrete = FALSE, pvalue = TRUE)
  res$stat
  res$pvalue
  
  ORD <- order(Y)
  Y2  <- array(0, n)
  Y2[ORD[1:floor(n / 3)]]                   <- 1
  Y2[ORD[(floor(n / 3) + 1):floor(2 * n / 3)]] <- 2
  Y2[ORD[(floor(2 * n / 3) + 1):n]]        <- 3
  
  res.2 <- gSNR(X, Y2, discrete = TRUE, pvalue = TRUE)
  res.2$stat
  res.2$pvalue

• BET: Binary Expansion Test – R package for binary expansion–based independence testing.
• LassoSIR – R package for sparse sliced inverse regression via Lasso.
• GroupTest – R package for multiple testing procedures with grouped hypotheses.
• CLAT: CDF and locfdr Assisted Testing – R package for CDF- and locfdr-assisted multiple testing.
  Show installation & example

  library(devtools)
  install_github("zhaozhg81/CLAT")

  library(CLAT)
  p <- 1000
  X <- c(rnorm(100, -2, 1),
         rnorm(100,  2, 1),
         rnorm(800,  0, 1))
  q <- 0.05
  
  ## One-sided test
  p.right    <- pnorm(X, lower.tail = FALSE)
  clat.right <- CLAT(p.right, q)
  rejind     <- which(clat.right$SigInd != 0)
  
  ## Two-sided test
  p.right    <- pnorm(X, lower.tail = FALSE)
  clat.right <- CLAT(p.right, q)
  p.left     <- pnorm(X, lower.tail = TRUE)
  clat.left  <- CLAT(p.left, q)
  rejind     <- c(which(clat.left$SigInd != 0),
                 which(clat.right$SigInd != 0))

• Bayesian LASSO with Zero Inflated Mixture Prior.
• EBVariant: An empirical Bayes testing procedure for detecting variants in analysis of next generation sequencing data.