QQNORM

# Apropos of the following posts:

# http://stats.stackexchange.com/q/94106/67822
# http://stats.stackexchange.com/a/183647/67822
# http://stats.stackexchange.com/a/101290/67822

# The presence of asymptotic values in a t distribution really messes up the histogram. This can be addressed by establishing
# limits to the x axis of the histogram, and increasing the number of breaks:

set.seed(99); n <- 1e4; y <- rt(n, df = 2); summary(y); 
hist(y, xlim=c(-5,5), breaks=2000)

# However, the qqnorm will still be rather distorted:

qqnorm(y)

# To address this problem, we can get rid of asymptotic values:

y <- y[y > quantile(y, .25) - 1.5 * IQR(y) &
       y < quantile(y, .75) + 1.5 * IQR(y)]
summary(y)
hist(y, xlim=c(-5,5), breaks=50)
qqnorm(y)