When lambda is large, we can use Normal(lambda, lambda) to approximate Poisson (This can be proved by using Continuity Theorem). The following is R code and output.
Pois2Norm <- function(lambda = 100, n=1000) {
x <- rpois(n, lambda)
minx <- min(x)
maxx <- max(x)
plot(density(x), main = "Use Normal to approximate Poisson when lambda is large", xlab = "X")
xnorm <- seq(from = minx, to = maxx, length = n)
lines(xnorm, dnorm(xnorm, mean = lambda, sd = sqrt(lambda)), col = "red")
x.pos <- minx + 0.01*(maxx - minx)
miny <- min(density(x)$y)
maxy <- max(density(x)$y)
y.pos <- maxy - 0.01*(maxy - miny)
legend(x.pos, y.pos, lty = c(1, 1), col = c("black", "red"), legend = c("Poisson", "Normal"))
}
When lambda = 10,
When lambda = 100,
When lambda = 1000,
When lambda is large, we can use Normal(lambda, lambda) to approximate Poisson (This can be proved by using Continuity Theorem). The following is R code and output.
When lambda = 10, When lambda = 100, When lambda = 1000,