DATA 352/551: Assignment 3

Central Limit Theorem simulation

Review

In class, we carried out the following simulation to understand the Central Limit Theorem.

Suppose you are teaching a class with n = 25 students. Simulate 25 exam scores from a uniform (0, 100) or normal (mean = 80, sd = 3) or Beta (shape1 = 6, shape2 = 2) distribution and calculate the class AVERAGE of your simulated class. Repeat for 10000 times, each time recording the class average. Plot a histogram of the 10000 simulated class averages and estimate the mean and standard deviation (sd) of this distribution.

Show code
nsim <- 10000
result <- rep(NA, nsim)

for(i in 1:nsim){
  exam_score <- runif(25, min = 0, max = 100)
  result[i] <- mean(exam_score)
}
hist(result)

mean(result)
[1] 49.9073
sd(result)
[1] 5.749079

Your results show that, regardless of your starting distribution (uniform, normal, or beta), the simulated class averages are very “well-behaved”: the histogram of the 10000 simulated averages should look normal (symmetric bell-curve) in each case.

Effect of sample size

Work in groups and submit your groups’ findings in this Google doc. Use screenshot (for code output or plots) as needed.

Design and run a simulation by repeating what you have done above but with different values of \(n\), as well as different starting distributions. For instance, you can start with:

  • different shapes of the Beta distribution (by changing the values of the shape1, shape2 parameters)
  • exponential distribution (try picking parameter values that would give a very skewed shape)
  • discrete distribution (for instance, imagine the exam scores can only be 100, 80, or 50; random sampling can be done by using sample instead of runif/rnorm/rbeta etc. )

Your group should try three or more different settings. In each case, as before, plot a histogram of the simulated averages to assess normality. In text, summarize your findings: what would you recommend as a sufficient sample size for the Central Limit Theorem to work in different settings? In introductory level statistics courses, we tell students to make sure that the sample size is at least 30. Do you think this is a reasonable cutoff? If someone asks, “my sample size is only 10, do you think I can run hypothesis tests that are based on the Central Limit Theorem?” What would you say?