Chapter 8
Sampling Distributions
Sampling distributions are probability distributions of statistics.
8.1 Distribution of the Sample Mean
Sampling distribution for random sample average, X̄, is described in this section. The
central limit theorem (CLT) tells us no matter what the original parent distribution,
sampling distribution of X̄ is typically normal when n ≥ 30. Related to this,
2
2 σX σX
µX̄ = µX , σX̄ = , σX̄ = √ .
n n
Exercise 8.1 (Distribution of the Sample Mean)
1. Practice with CLT: average, X̄.
(a) Number of burgers.
Number of burgers, X, made per minute at Best Burger averages µX = 2.7
burgers with a standard deviation of σX = 0.64 of a burger. Consider
average number of burgers made over random n = 35 minutes during day.
i. µX̄ = µX = (circle one) 2.7 / 2.8 / 2.9.
σX 0.64
ii. σX̄ = √ n
=√ 35
= 0.10817975 / 0.1110032 / 0.13099923.
iii. P X̄ > 2.75 ≈ (circle one) 0.30 / 0.32 / 0.35.
(Stat, Calculators, Normal, Mean: 2.7, Std. Dev.: 0.10817975,
Prob(X ≥ 2.75) = ? , Compute.)
iv. P 2.65 < X̄ < 2.75 ≈ (circle one) 0.36 / 0.39 / 0.45.
(Stat, Calculators, Normal, Between, Mean: 2.7, Std. Dev.: 0.10817975, Prob(2.65 ≤ X ≤ 2.75))
(b) Temperatures.
Temperature, X, on any given day during winter in Laporte averages µX =
135
136 Chapter 8. Sampling Distributions (Lecture Notes 8)
0 degrees with standard deviation of σX = 1 degree. Consider average
temperature over random n = 40 days during winter.
i. µX̄ = µX = (circle one) 0 / 1 / 2.
σX
ii. σX̄ = √ n
= √140 = 0.0900234 / 0.15811388 / 0.23198455.
iii. P X̄ > 0.2 ≈ (circle one) 0.03 / 0.10 / 0.15.
(Stat, Calculators, Normal, Mean: 0, Std. Dev.: 0.15811388,
Prob(X > 0.2) = ? , Compute.)
iv. P X̄ > 0.3 ≈ (circle one) 0.03 / 0.10 / 0.15.
(Stat, Calculators, Normal, Mean: 0, Std. Dev.: 0.15811388,
Prob(X > 0.3) = ? , Compute.)
Since P X̄ > 0.3 ≈ 0.03 < 0.05, 0.3o is / is not unusual.
(c) Another example.
Suppose X has distribution where µX = 1.7 and σX = 1.5.
i. µX̄ = µX = (circle one) 2.3 / 1.7 / 2.4.
σX
ii. σX̄ = √ n
= √1.5
49
= 0.0243892 / 0.14444398 / 0.21428572.
iii. If n = 49, P (−2 < X̄ < 2.75) ≈ (circle one) 0.58 / 0.86 / 0.999.
(Stat, Calculators, Normal, Between, Mean: 1.7, Std. Dev.: 0.21428572, Prob(-2 ≤ X ≤ 2.75))
iv. True / False.
If n = 15, P (−2 < X̄ < 2.75) cannot be calculated since n = 15 < 30.
σX
v. σX̄ = √ n
= √1.5
15
= 0.0243892 / 0.14444398 / 0.38729835.
vi. If n = 15 and normal, P (−2 < X̄ < 2.75) ≈ 0.75 / 0.78 / 0.997.
(Stat, Calculators, Normal, Between, Mean: 1.7, Std. Dev.: 0.38729835, Prob(-2 ≤ X ≤ 2.75))
(d) Dice average.
What is the
chance, in n = 30 rolls of a fair die, average is between 3.3
and 3.7, P 3.3 < X̄ < 3.7 ? What if n = 3?
1
i. µX̄ = µX = 1 6
+ · · · + 6 16 = (circle one) 2.3 / 3.5 / 4.3.
(Blank data table. Relabel var1 die, var2 frequency. Type 1, 2, 3, 4, 5, 6, in die column, and 1, 1,
1, 1, 1, 1 in frequency column. Data, Save Data, 8.1 die distribution. Dat, compute expression,
Expression: frequency/6, New column name: P(x), Compute. Stat, Calculators, Custom, Values
in: die, Weights in: P(x), Okay. Notice, Mean: 3, Std. Dev.: 1.7078252.)
r
1 1
ii. σX = (1 − 3.5)2 6
+ · · · + (6 − 3.5)2 6
=
1.7078252 / 2.131145 / 3.3409334.
σX 1.71
iii. If n = 30, σX̄ = √ n
=√ 30
= (circle one) 0.31 / 0.75 / 1.14.
iv. If n = 30, P 3.3 < X̄ < 3.7 ≈ (circle one) 0 / 0.20 / 0.48.
(Stat, Calculators, Normal, Between, Mean: 3.5, Std. Dev.: 0.3118048, Prob(3.3 ≤ X ≤ 3.7))
v. True / False.
If n = 3, P 3.3 < X̄ < 3.7 cannot be calculated because n = 3 < 30.
Section 1. Distribution of the Sample Mean (Lecture Notes 8) 137
2. Understanding CLT: Montana fishing trip.
(a) Sampling distributions of average, n = 1, 2, 3.
0.40 0.40
σ = 0.75
0.20
(a)
_________
0.20
1 2 3 X1
= 1 2 3 X1
1
0.40 0.40
0.40
0.20 0.20
σ = 0.53
+
(b) _________
1 2 3 X1 1 2 3 X2
=
0.20
1 3/2 2 5/2 3 _ X + X2
2 X=
_____
1
µ = 1.8
0.40 0.40 0.40
0.40
0.20 0.20 0.20
(c) _________ =
1 2 3 X1
+
1 2 3 X2
+
1 2 3 X3
0.20
1 4_ _
5 2 7_ 8_ 3
σ = 0.43
_ X + X2+ X3
3 3 3 3 3
X=
_____
1
Figure 8.1 (Comparing sampling distributions of sample mean)
As random sample size, n, increases, sampling distribution of average, X̄,
changes shape and becomes more (circle one)
i. rectangular–shaped.
ii. bell–shaped.
iii. triangular–shaped.
Central limit theorem (CLT) says no matter what the original parent distri-
bution, sampling distribution of average is typically normal when n > 30.
(b) In addition to sampling distribution becoming more normal–shaped as
random sample size increases, mean of average, µX̄ = 1.8 (circle one)
σ2
i. decreases and is equal to nX ,
ii. remains same and is equal to µX = 1.8,
iii. increases and is equal to nµX ,
138 Chapter 8. Sampling Distributions (Lecture Notes 8)
and standard deviation of average, σX̄ (circle one)
σX
i. decreases and is equal to √ n
.
ii. remains same and is equal to σX .
iii. increases and is equal to nσX .
(c) After n = 30 trips to lake, sampling distribution in average number of fish
caught is essentially normal (why?) where
µX̄ = µX = (circle one) 1.2 / 1.5 / 1.8,
0.75
σX̄ = √ 30
≈ 0.12677313 / 0.13693064 / 0.2449987,
(Data, Compute expression, Expression: 0.75/sqrt(30), Compute.)
and chance average number of fish is less than 1.95 is
P (X̄ < 1.95) ≈ (circle one) 0.73 / 0.86 / 0.94.
(Stat, Calculators, Normal, Mean: 1.8, Std. Dev.: 0.13693064,
Prob(X < 1.95) = ? , Compute.)
(d) After n = 35 trips to lake, sampling distribution in average number of fish
caught is essentially normal where
µX̄ = µX = (circle one) 1.2 / 1.5 / 1.8,
0.75
σX̄ = √ 35
≈ 0.12677313 / 0.13693064 / 0.2449987,
(Data, Compute expression, Expression: 0.75/sqrt(35), Compute.)
and chance average number of fish is less than 1.95 is
P (X̄ < 1.95) ≈ (circle one) 0.73 / 0.88 / 0.94.
(Stat, Calculators, Normal, Mean: 1.8, Std. Dev.: 0.12677313,
Prob(X < 1.95) = ? , Compute.)
(e) Chance average number of fish is less than 1.95 after 30 trips, P (X̄ <
1.95) ≈ 0.86, is smaller than / larger than chance average number of
fish is less than 1.95 after 35 trips, P (X̄ < 1.95) ≈ 0.88.
average number of fish caught,
n = 30 trips,
chance = 0.86 mean 1.8, SD 0.14
1.38 1.52 1.66 1.80 1.94 2.08 2.22
1.95
average number of fish caught,
n = 35 trips,
chance = 0.88 mean 1.8, SD 0.13
1.38 1.52 1.66 1.80 1.94 2.08 2.22
Section 1. Distribution of the Sample Mean (Lecture Notes 8) 139
Figure 8.2 (Chance when n = 30 compared to chance when n = 35)
(f) The CLT is useful because (circle one or more):
i. No matter what original parent distribution is, as long as a large
enough random sample is taken, average of this sample follows a nor-
mal distribution.
ii. In practical situations where it is not known what parent probability
distribution to use, as long as a large enough random sample is taken,
average of this sample follows a normal distribution.
iii. Rather than having to deal with many different probability distribu-
tions, as long as a large enough random sample is taken, average of
this sample follows one distribution, normal distribution.
iv. Many distributions in statistics rely in one way or another on normal
distribution because of CLT.
(g) True / False Central limit theorem requires not only n ≥ 30, but also a
random sample of size n ≥ 30 is used.
3. Simulating sampling distribution for average1: Montana fishing trip.
1
Text uses simulations, rather than exact sampling distributions, in discussion of CLT.
140 Chapter 8. Sampling Distributions (Lecture Notes 8)
Exact Sampling Distributions Simulated Sampling Distributions
4
s = 0.876
2 10 simulations
0.40
σ = 0.75 1 3
_ 2
(a) 0.20 x = 2.1
X1
40
1 2 3 s = 0.744
X1 20 100 simulations
n=1
1 2 3 X1
_
x = 1.82
µ = 1.8 12
0.40 6
s_= 0.510
x
30 simulations
0.20
σ_= 0.53
X 1 3/2 2 5/2 3 _
(b) _
x = 1.92
X
120
1 3/2 2 5/2 3 _ X + X2
_____
1
X=
2 60
s_= 0.530
x 300 simulations
n=2
_ 2 5/2 3
1 3/2 _
X
x = 1.8
12
6 s_= 0.366
0.40 x 30 simulations
1 4
_ 5
_ 2 7
_ 8
_ 3 _
0.20 σ_= 0.43
(c)
3 3 3 3
_ X
X x = 1.877
120
1 4
_ 5
_ 2 7
_ 8
_ 3 _ X + X2+ X3
_____
1
3 3 3 3
X= 3 60 s_= 0.406
x 300 simulations
n=3 1 4
_ 5
_ 2 7
_ 8
_ 3 _
3 3 3 3
_ X
x = 1.767
Figure 8.3 (Simulating sampling distributions: Montana fishing trip.)
(a) Figure 8.4(a). Parent distribution, n = 1.
Approximate simulated histogram shape and (x̄, s) all (choose one)
improve, are typically closer to
worsen, are typically farther away from
parent distribution shape and (µ, σ) ≈ (1.8, 0.75).
as number of simulations increases, from 10 to 100.
(b) Figure 8.4(b). Sampling distribution average, n = 2.
Approximate simulated histogram shape and (x̄, s) all (choose one)
improve, are typically closer to
worsen, are typically farther away from
parent distribution shape and (µ, σ) ≈ (1.8, 0.53).
as number of simulations increases, from 30 to 300.
(c) Figure 8.4(c). Sampling distribution average, n = 3.
Approximate simulated histogram shape and (x̄, s) all (choose one)
Section 2. Distribution of the Sample Proportion (Lecture Notes 8) 141
improve, are typically closer to
worsen, are typically farther away from
parent distribution shape and (µ, σ) ≈ (1.8, 0.43).
as number of simulations increases, from 10 to 100.
(d) Sample Size Versus Number of Simulations.
As number trips to lake (sample size) increases, n = 1 to n = 3, sampling
distribution of average does / does not become more normal.
As number of simulations increase, approximate sampling distribution
does / does not become more normal unless distribution normal.
8.2 Distribution of the Sample Proportion
Central limit theorem (CLT) tells us no matter what the original parent distribution,
sampling distribution of random2 sample proportion3 , p̂ = Xn , is typically normal
when np(1 − p) ≥ 10 and n ≤ 0.05N. Related to this,
s
p(1 − p)
µp̂ = p, σp̂ = .
n
Exercise 8.2 (Distribution of the Sample Proportion)
1. Proportion of wins.
Lawyer estimates she wins 40% of her cases (p = 0.4), and currently represents
n = 50 defendants. Let X represent number of wins (of 50 cases) and so p̂ = Xn
proportion of wins (of 50 cases). Use CLT to approximate chance she wins at
least one–half of her cases,
1
P p̂ > = P (p̂ > 0.5) .
2
(a) Check assumptions.
Since np(1 − p) = 50(0.4)(1 − 0.4) = 12 ≥ 10, assumptions necessary to
proceed with approximation are (choose one) satisfied / violated.
(b) µp̂ = p = (circle one) 0.3 / 0.4 / 0.5.
q q
(c) σp̂ = p(1−p)
n
= 0.4(1−0.4)
50
≈ 0.0011655 / 0.03855667 / 0.06928203.
(Data, Compute expression, Expression: sqrt(0.4*0.6/50), Compute.)
2
If population finite, simple random sample size must be small, no more than 5% of population
n
size, N ≤ 0.05, to ensure independence of items in sample.
3
Previously discussed normal approximation to binomial is an example of CLT. We look at this
approximation again, only focus on sample proportion, p̂ = X n , rather than sample number, X.
142 Chapter 8. Sampling Distributions (Lecture Notes 8)
(d) P (p̂ > 0.5) ≈ (circle one) 0.07 / 0.11 / 0.13.
(Stat, Calculators, Normal, Mean: 0.4, Std. Dev.: 0.06928203,
Prob(X ≥ 0.5) = ? , Compute.)
Since P (p̂ > 0.5) ≈ 0.07 > 0.05, p̂ = 0.5 is typical / unusual.
2. Another example.
Let p = 0.63 and n = 45 and approximate
P (p̂ < 0.41) .
(a) Check assumptions.
Since np(1−p) = 45(0.63)(1−0.63) = 10.4895 > 10, assumptions necessary
to proceed with approximation are (choose one) satisfied / violated.
(b) µp̂ = p = (circle one) 0.54 / 0.60 / 0.63.
q q
(c) σp̂ = p(1−p)
n
= 0.63(1−0.63)
45
≈ 0.06177628 / 0.07197221 / 0.08900453.
(Data, Compute expression, Expression: sqrt(0.63*0.37/45), Compute.)
(d) P (p̂ < 0.41) ≈ (circle one) 0.001 / 0.002 / 0.003.
(Stat, Calculators, Normal, Mean: 0.63, Std. Dev.: 0.07197221,
Prob(X ≤ 0.41) = ? , Compute.)
Since P (p̂ < 0.41) ≈ 0.001 < 0.05, p̂ = 0.63 is typical / unusual.
3. And another example.
Let p = 0.25 and n = 42, N = 10000 and approximate
P (0.22 < p̂ < 0.28) .
(a) Check assumptions.
Since np(1 − p) = 42(0.25)(0.75) = 7.875 < 10 and n = 42 < 0.05N =
0.05(10000) = 500, assumptions necessary to proceed with approximation
are satisfied / violated. So we won’t.
4. Proportion of wins again.
Lawyer estimates she wins 40% of her cases (p = 0.4), and currently represents
n = 50 defendants. What is the chance4 she wins at least 30 cases?
(a) µp̂ = p = (circle one) 0.3 / 0.4 / 0.5.
q q
(b) σp̂ = p(1−p)
n
= 0.4(1−0.4)
50
≈ 0.00123234 / 0.0387786 / 0.06928203.
(Data, Compute expression, Expression: sqrt(0.4*0.6/50), Compute.)
30
(c) P (X ≥ 30) = P p̂ ≥ 50
≈ (circle one) 0.001 / 0.002 / 0.003.
(Stat, Calculators, Normal, Mean: 0.4, Std. Dev.: 0.06928203,
Prob(X ≥ 0.6) = ? , Compute.)
4
Ignore, do not use, the continuity correction here.