In PowerUpR: Power Analysis Tools for Multilevel Randomized Experiments

knitr::opts_chunk$set(echo = TRUE)
library(PowerUpR)

To install and load PowerUpR:

install.packages("PowerUpR")
library(PowerUpR)

PowerUpR functions are not vectorized to avoid possible complications. However, researchers often explore variety of design charactersitics when they determine their sample size. Creating custom plots and tables may assist with their decision. In what follows, example code snippets are provided to demonstrate vectorization of PowerUpR functions over single or multiple design parameters.

NOTE: We would like to thank Dr. Andi Fugard for inspiring this vignette. Dr. Fugard realized that PowerUpR functions does not evaluate arguments when they are embedded within a user-specificed function. This vignette provides example vectorization of PowerUpR functions for creating custom plots and tables.

Plotting MDES (mdes) against level-3 intra-class correlation coefficient (rho3)

custom_fun <- function(x) {
  parms <- list(rho3 = x,
                power = .80, rho2 = .06,
                g3 = 1, r21 = .55, r22 = .50, r23 = .45,
                p = .40, n = 10, J = 2, K = 83)
  design <- do.call("mdes.cra3", parms)
  design$mdes[1]
}

x = seq(.10,.90,.01)
mdes <- mapply(custom_fun, x)
plot(x, mdes, type = "l", xlab = "rho3")

Plotting power (power) against sample size (K) and explanatory power of level-3 covarites (r23)

custom_fun <- function(x, y) {
  parms <- list(K = x, r23 = y, 
                es = .23, rho2 = .06, rho3 = .18,
                g3 = 1, r21 = .55, r22 = .50,
                p = .40, n = 10, J = 2)
  design <- do.call("power.cra3", parms)
  design$power
}

x = seq(10,100,5)
power.r23.30 <- mapply(custom_fun, x, .30)
power.r23.40 <- mapply(custom_fun, x, .40)
power.r23.50 <- mapply(custom_fun, x, .50)
power.r23.60 <- mapply(custom_fun, x, .60)

# plot
plot(x, power.r23.30, pch = 18, type = "b",
     ylim = c(0,1),  xlab = "K", ylab = "Power")
lines(x, power.r23.40, col = 2, pch = 19, type = "b")
lines(x, power.r23.50, col = 3, pch = 20, type = "b")
lines(x, power.r23.60, col = 4, pch = 21, type = "b")
legend("bottomright",  bty = "n",
       legend = c("r23=.30", "r23=.40", "r23=.50", "r23=.60"), 
       col = c(1, 2, 3, 4), lty = c(1, 1, 1, 1), pch = c(18, 19, 20, 21))
grid(nx = 20, ny = 18)

Table of required sample size (K) for various effect size (es) values

custom_fun <- function(x) {
  parms <- list(es = x, power = .80, rho2 = .06, rho3 = .18,
                g3 = 1, r21 = .55, r22 = .50, r23 = .45,
                p = .40, n = 10, J = 2)
  design <- do.call("mrss.cra3", parms)
  design$K
}

x = seq(.10,.50,.05)
K <- mapply(custom_fun, x)
table <- data.frame(es = x, K = K)

print(table)

Table of required sample size (K) for various effect size (es) and R-squared values (r23) values

custom_fun <- function(x1,x2) {
  parms <- list(es = x1, r23 = x2,
                power = .80, rho2 = .06, rho3 = .18,
                g3 = 1, r21 = .55, r22 = .50,
                p = .40, n = 10, J = 2)
  design <- do.call("mrss.cra3", parms)
  design$K
}

vec.custom_fun <- Vectorize(custom_fun, c("x1", "x2"))

x1 = seq(.10,.50,.05)
x2 = seq(.20,.70,.10)

table.K <- outer(x1, x2, vec.custom_fun)
rownames(table.K) <- paste0("es=",x1)
colnames(table.K) <- paste0("r23=",x2)