PROBLEM SET 2: SAMPLING DISTRIBUTION

In this problem set, we aim to empirically construct the sampling distribution of the mean,
with the goal of understanding the Central Limit Theorem.

Part I: Construct a Population

A deck has three cards with three different point values: 0, 1, and 3. We will imagine
drawing from this deck one billion times. The set of all possible values forms a population. Of
course, we cannot observe all data, and thus the population mean and variance are fixed but
unknown. However, we can model the population using probability theory.
First, we define
X = the value when we draw a card.
Then the population is modeled by the random variable X. Now you will construct the distri-
bution table (fill in the blanks):
X 0 1 3
P
From the table distribution, you can compute the population mean, variance and standard
deviation (fill in the blanks)
• Population mean µ = . . .

• Population variance σ 2 = . . .

• Population standard deviation σ = . . .

Solution

Assume each card is equally likely: P (X = 0) = P (X = 1) = P (X = 3) = 31 .

X 0 1 3
1 1 1
P
3 3 3

0+1+3 4 02 + 1 2 + 3 2 10
µ = E[X] = = , E[X 2 ] = = .
3 3 3 3
( )2 √ √
10 4 30 − 16 14 14 14
σ = E[X ] − µ =
2 2 2
− = = , σ= = .
3 3 9 9 9 3
(Numerically: µ ≈ 1.333, σ 2 ≈ 1.556, σ ≈ 1.247.)

Part II: Sampling distribution of the mean of the sample size 2

In this part, we collect all random sample of the sample size 2 and, then, construct the
distribution for the sample mean. Now, define
X1 = the first value we draw 2 times.
X2 = the first value we draw 2 times.

1
Therefore, the sample mean is
1
X = (X1 + X2 ).
2
You will determine the distribution table of X (fill in the blanks).
X̄ Ordered outcomes Count Probability
(0, 0)
(0, 1), (1, 0)
(1, 1)
(0, 3), (3, 0)
(1, 3), (3, 1)
(3, 3)
Next, answer the following questions:
• X and X have the same distribution (True/False): ….
• Find P (1 ≤ X ≤ 1.5) = . . .
• The mean of X and X are the same (True/False): …
• Sketch the histogram of the sampling distribution of X:
Solution

Interpret X1 as the first draw and X2 as the second draw; draws are independent with
P = 1/3 each outcome. There are 3 × 3 = 9 equally likely ordered pairs, each with
probability 1/9.
Possible sample means and probabilities:

X̄ Ordered outcomes Count Probability

1
0 (0, 0) 1 9
2
0.5 (0, 1), (1, 0) 2 9
1
1 (1, 1) 1 9
2
1.5 (0, 3), (3, 0) 2 9
2
2 (1, 3), (3, 1) 2 9
1
3 (3, 3) 1 9

Answers:
• X and X have the same distribution: False (supports differ).

• P (1 ≤ X ≤ 1.5) = P (X = 1) + P (X = 1.5) = 1
9
+ 2
9
= 1
3
.

• E[X] = 12 (E[X1 ] + E[X2 ]) = µ, so the means are the same: True .

Probability
2
9
1
9
0
• 0 0.5 1 1.5 2 3 X̄

Histogram of the sampling distribution of X̄ for n = 2.

2
 Part III: Sampling distribution of the sample mean for a large sample size

By the same fashion above and computer aid, we can construct the distribution of the sample
mean X when the sample size n is large. In particular, the following histograms illustrates cases
n = 50, 100, 1000. From the histograms above, answer the following questions:

Figure 1: Histograms of the sampling distribution of X̄ for n = 50, 100, 1000.

• When the sample size increases, the distribution of the sample mean becomes increasingly
(bell-shaped / chaotic / uniform): …

• When the sample size increases, the central value of the histograms remains the same
(True / False): …

• When the sample size increases, the histogram becomes (narrower / wider): …. Thus, the
variance (increases / decreases): …

• When the sample size increases, P (1 ≤ X ≤ 1.5) (increases / decreases): …. Therefore,

4
sample means are more likely to be (far / close): … to the population mean .
3
Solution

• bell-shaped — X tends to normality as n grows.

4
• True — E[X] = µ = 3
for all n .
14
• narrower , hence variance decreases (since Var(X) = σ 2 /n where σ 2 = 9
).

• increases — more mass concentrates near µ; thus sample means are more likely
to be close to µ = 43 .