Generating data when the tilting parameter is of class integer or numeric return different results. See the example below, where cc varies from 1 to 24 and draws from the Polya-Gamma are acquired setting z = cc and z = as.numeric(cc).
B <- 10000
x1 <- matrix(NA, nrow = B, ncol = length(z_vals))
x2 <- x1
set.seed(1234)
for (bb in 1 : B) {
for (cc in 1 : length(z_vals)) {
var1 <- cc
var2 <- as.numeric(cc)
if (var1 != var2) {
print('var1 and var2 not the same')
}
x1[bb, cc] <- BayesLogit::rpg(1, h = 1, z = var1)
x2[bb, cc] <- BayesLogit::rpg(1, h = 1, z = var2)
}
}
par(mfrow = c(2, 3))
for (cc in 1 : length(z_vals)) {
plot((density(x1[!is.na(x1[, cc]), cc])), main = cc)
lines(density(x2[!is.na(x2[, cc]), cc], na.rm = TRUE), col = 'red')
}
Generating data when the tilting parameter is of class integer or numeric return different results. See the example below, where
ccvaries from 1 to 24 and draws from the Polya-Gamma are acquired settingz = ccandz = as.numeric(cc).