Normal Distribution
Michael Foley
January 7, 2019
R Functions dnorm, pnorm, qnorm, and rnorm
Random variable \(X\) is distributed \(X \sim N(\mu, \sigma^2)\) if \(f(X)=\frac{{1}}{{\sigma \sqrt{{2\pi}}}}e^{-.5(\frac{{x-\mu}}{{\sigma}})^2}\).
R function dnorm(x, mean, sd) is the probability of x when the mean is mean and the standard deviation is sd. R function pnorm(q, mean, sd, lower.tail) is the cumulative probability (lower.tail = TRUE for left tail, lower.tail = FALSE for right tail) of less than or equal to value q. R function rnorm(n, mean, sd) returns n random numbers from the normal distribution X~N(mu, sigma^2). R function qnorm(p, mean, sd, lower.tail) is the value of x at the qth percentile (lower.tail = TRUE).
Example
IQ scores are distributed \(X \sim N(100, 16^2\). What is the probability a randomly selected person’s IQ is <90?
my_mean = 100
my_sd = 16
my_x = 90
# exact
pnorm(q = my_x, mean = my_mean, sd = my_sd, lower.tail = TRUE)## [1] 0.2659855# simulated
mean(rnorm(n = 10000, mean = my_mean, sd = my_sd) <= my_x)## [1] 0.2724library(dplyr)
library(ggplot2)
data.frame(x = 0:1500 / 10,
prob = pnorm(q = 0:1500 / 10,
mean = my_mean,
sd = my_sd,
lower.tail = TRUE)) %>%
mutate(cdf = ifelse(x > 0 & x <= my_x, prob, 0)) %>%
ggplot() +
geom_line(aes(x = x, y = prob)) +
geom_area(aes(x = x, y = cdf), alpha = 0.3) +
labs(title = bquote('X~N('~mu==.(my_mean)~','~sigma^{2}==.(my_sd)^{2}~')'),
subtitle = bquote('P(X<='~.(my_x)~') when mean is'~.(my_mean)~' and variance is'~.(my_sd)^{2}~'.'),
x = "x",
y = "Probability") 
Example
IQ scores are distributed \(X \sim N(100, 16^2\). What is the probability a randomly selected person’s IQ is >140?
my_mean = 100
my_sd = 16
my_x = 140
# exact
pnorm(q = my_x, mean = my_mean, sd = my_sd, lower.tail = FALSE)## [1] 0.006209665# simulated
mean(rnorm(n = 10000, mean = my_mean, sd = my_sd) > my_x)## [1] 0.0049library(dplyr)
library(ggplot2)
data.frame(x = 0:1500 / 10,
prob = pnorm(q = 0:1500 / 10,
mean = my_mean,
sd = my_sd,
lower.tail = TRUE)) %>%
mutate(cdf = ifelse(x > my_x & x < 1000, prob, 0)) %>%
ggplot() +
geom_line(aes(x = x, y = prob)) +
geom_area(aes(x = x, y = cdf), alpha = 0.3) +
labs(title = bquote('X~N('~mu==.(my_mean)~','~sigma^{2}==.(my_sd)^{2}~')'),
subtitle = bquote('P(X<='~.(my_x)~') when mean is'~.(my_mean)~' and variance is'~.(my_sd)^{2}~'.'),
x = "x",
y = "Probability") 
Example
IQ scores are distributed \(X \sim N(100, 16^2\). What is the probability a randomly selected person’s IQ is between 92 and 114?
my_mean = 100
my_sd = 16
my_x_l = 92
my_x_h = 114
# exact
pnorm(q = my_x_h, mean = my_mean, sd = my_sd, lower.tail = TRUE) -
pnorm(q = my_x_l, mean = my_mean, sd = my_sd, lower.tail = TRUE)## [1] 0.5006755library(dplyr)
library(ggplot2)
data.frame(x = 0:1500 / 10,
prob = pnorm(q = 0:1500 / 10,
mean = my_mean,
sd = my_sd,
lower.tail = TRUE)) %>%
mutate(cdf = ifelse(x > my_x_l & x <= my_x_h, prob, 0)) %>%
ggplot() +
geom_line(aes(x = x, y = prob)) +
geom_area(aes(x = x, y = cdf), alpha = 0.3) +
labs(title = bquote('X~N('~mu==.(my_mean)~','~sigma^{2}==.(my_sd)^{2}~')'),
subtitle = bquote('P(X<='~.(my_x)~') when mean is'~.(my_mean)~' and variance is'~.(my_sd)^{2}~'.'),
x = "x",
y = "Probability") 
Example
Class scores are distributed \(X \sim N(70, 10^2\). If the instructor wants to give A’s to >=85th percentile and B’s to 75th-85th percentile, what are the cutoffs?
my_mean = 70
my_sd = 10
my_pct_l = .75
my_pct_h = .85
qnorm(p = my_pct_l, mean = my_mean, sd = my_sd, lower.tail = TRUE)## [1] 76.7449qnorm(p = my_pct_h, mean = my_mean, sd = my_sd, lower.tail = TRUE)## [1] 80.36433library(dplyr)
library(ggplot2)
data.frame(x = 0:1000 / 10,
prob = pnorm(q = 0:1000 / 10,
mean = my_mean,
sd = my_sd,
lower.tail = TRUE)) %>%
mutate(cdf = ifelse(prob > my_pct_l & prob <= my_pct_h, prob, 0)) %>%
ggplot() +
geom_line(aes(x = x, y = prob)) +
geom_area(aes(x = x, y = cdf), alpha = 0.3) +
labs(title = bquote('X~N('~mu==.(my_mean)~','~sigma^{2}==.(my_sd)^{2}~')'),
subtitle = bquote('P(X<=x) = ['~.(my_pct_l)~','~.(my_pct_h)~'] when mean is'~.(my_mean)~' and variance is'~.(my_sd)^{2}~'.'),
x = "x",
y = "Probability") 
Normal Approximation to Binomial
The CLT implies that certain distributions can be approximated by the normal distribution.
The binomial distribution \(X \sim B(n,p)\) is approximately normal with mean \(\mu = n p\) and variance \(\sigma^2=np(1-p)\). The approximation is useful when the expected number of successes and failures is at least 5: \(np>=5\) and \(n(1-p)>=5\).
Example
A measure requires p>=50% popular to pass. A sample of n=1,000 yields x=460 approvals. What is the probability that the overall population approves, P(X)>0.5?
my_x = 460
my_p = 0.50
my_n = 1000
my_mean = my_p * my_n
my_sd = round(sqrt(my_n * my_p * (1 - my_p)), 1)
# Exact binomial
pbinom(q = my_x, size = my_n, prob = my_p, lower.tail = TRUE)## [1] 0.006222073# Normal approximation
pnorm(q = my_x, mean = my_p * my_n, sd = sqrt(my_n * my_p * (1 - my_p)), lower.tail = TRUE)## [1] 0.005706018library(dplyr)
library(ggplot2)
library(tidyr)
data.frame(x = 400:600,
Normal = pnorm(q = 400:600,
mean = my_p * my_n,
sd = sqrt(my_n * my_p * (1 - my_p)),
lower.tail = TRUE),
Binomial = pbinom(q = 400:600,
size = my_n,
prob = my_p,
lower.tail = TRUE)) %>%
gather(key = "Distribution", value = "cdf", c(-x)) %>%
ggplot(aes(x = x, y = cdf, color = Distribution)) +
geom_line() +
labs(title = bquote('X~B(n='~.(my_n)~', p='~.(my_p)~'), '~'X~N('~mu==.(my_mean)~','~sigma^{2}==.(my_sd)^{2}~')'),
subtitle = "Normal approximation to the binomial",
x = "x",
y = "Probability") 
The Poisson distribution \(x~P(\lambda)\) is approximately normal with mean \(\mu = \lambda\) and variance \(\sigma^2 = \lambda\), for large values of \(\lambda\).
Example
The annual number of earthquakes registering at least 2.5 on the Richter Scale and having an epicenter within 40 miles of downtown Memphis follows a Poisson distribution with mean \(\lambda=6.5\). What is the probability that at least \(x>=9\)* such earthquakes will strike next year?*
my_x = 9
my_lambda = 6.5
my_sd = round(sqrt(my_lambda), 2)
# Exact Poisson
ppois(q = my_x - 1, lambda = my_lambda, lower.tail = FALSE)## [1] 0.208427# Normal approximation
pnorm(q = my_x - 0.5, mean = my_lambda, sd = my_sd, lower.tail = FALSE)## [1] 0.216428library(dplyr)
library(ggplot2)
library(tidyr)
data.frame(x = 0:200 / 10,
Normal = pnorm(q = 0:200 / 10,
mean = my_lambda,
sd = my_sd,
lower.tail = TRUE),
Poisson = ppois(q = 0:200 / 10,
lambda = my_lambda,
lower.tail = TRUE)) %>%
gather(key = "Distribution", value = "cdf", c(-x)) %>%
ggplot(aes(x = x, y = cdf, color = Distribution)) +
geom_line() +
labs(title = bquote('X~P('~lambda~'='~.(my_lambda)~'), '~'X~N('~mu==.(my_lambda)~','~sigma^{2}==.(my_lambda)~')'),
subtitle = "Normal approximation to the Poisson",
x = "x",
y = "Probability") 
From Sample to Population
Suppose a person’s blood pressure typically measures 160±20 mm. If one takes n=5 blood pressure readings, what is the probability the average will be <=150?
my_mu = 160
my_sigma = 20
my_n = 5
my_x = 150
my_se = round(my_sigma / sqrt(my_n), 1)
pnorm(q = my_x, mean = my_mu, sd = my_sigma / sqrt(my_n), lower.tail = TRUE)## [1] 0.1317762library(dplyr)
library(ggplot2)
data.frame(x = 1000:2000 / 10,
prob = pnorm(q = 1000:2000 / 10,
mean = my_mu,
sd = my_sigma / sqrt(my_n),
lower.tail = TRUE)) %>%
mutate(cdf = ifelse(x > 0 & x <= my_x, prob, 0)) %>%
ggplot() +
geom_line(aes(x = x, y = prob)) +
geom_area(aes(x = x, y = cdf), alpha = 0.3) +
labs(title = bquote('X~N('~mu==.(my_mu)~','~sigma^{2}==.(my_se)^{2}~')'),
subtitle = bquote('P(X<='~.(my_x)~') when mean is'~.(my_mu)~' and variance is'~sigma~'/sqrt(n)'~.(my_se)^{2}~'.'),
x = "x",
y = "Probability") 