MTH-494 Sampling and Design Analysis

Lecture 09
Stratified Random Sampling
• Subgroups in the population are homogeneous
• Between groups there is heterogeneity.
• The strata should define a partition of the population. That is, it
should be collectively exhaustive and mutually exclusive: every
element in the population must be assigned to one and only one
stratum.
Mutually Exclusive Events: A1 ∩ A2∩ A3,….,∩ Ak = ∅
Collectively Exhaustive Events: A1 U A2U A3,….,U Ak = S
Stratified Random Sampling (Example 2.2)
Example 2.2, U={1, 2, 4, 4, 7, 7, 7 ,8}

• Consider the hypothetical population below (this population is also

used in Example 2.2). Consider the stratification below, with N1=N2=4.
The population is:

Consider the stratified sampling design in which n1=n2=2.

Stratified Random Sampling
• Write out all possible SRSs of size 2 from stratum 1, and find the
probability of each sample. Do the same for stratum 2
Stratum 1: 1, 2, 4, 8
Stratum 2: 4, 7, 7, 7

Example 2.2,
U={1, 2, 4, 4, 7, 7, 7 ,8}
Stratified Random Sampling
• Write out all possible SRSs of size 2 from stratum 1, and find the
probability of each sample. Do the same for stratum 2.
• Probability
P(Samples in S1)=1/6
P(Samples in S2)=1/6
P(Sample)= 1/NCn
Example 2.2,
U={1, 2, 4, 4, 4, 7, 7 ,8}
Stratified Random Sampling
• Using your work in (a), find the sampling distribution of 𝑡Ƹ𝑠𝑡𝑟 .
Samples ഥ 𝟏)
Means (𝒀 ഥ 1)
f(𝒀 Samples ഥ 𝟐)
Means (𝒀 ഥ 2)
f(𝒀
1,2 1.5 1/6 4,7 5.5 1/6
1,4 2.5 1/6 4,7 5.5 1/6
1,8 4.5 1/6 4,7 5.5 1/6
2,4 3 1/6 7,7 7 1/6
2,8 5 1/6 7,7 7 1/6
4,8 6 1/6 7,7 7 1/6
ഥ 𝟐)
Means (𝒀 ഥ 2)
f(𝒀
5.5 3/6
7.0 3/6
Stratified Random Sampling
• Using your work in (a), find the sampling distribution of 𝑡Ƹ𝑠𝑡𝑟 .
Samples ഥ 𝟏)
Total (𝒕ො1=N1𝒀 f(𝒕ො1) Samples ഥ 𝟐)
Total (𝒕ො2=N2𝒀 ഥ 𝟐)
f(𝒀
1,2 6 1/6 4,7 22 1/6
1,4 10 1/6 4,7 22 1/6
1,8 18 1/6 4,7 22 1/6
2,4 12 1/6 7,7 28 1/6
2,8 20 1/6 7,7 28 1/6
4,8 24 1/6 7,7 28 1/6
ഥ 𝟐)
Means (𝒀 ഥ 2)
f(𝒀
5.5 3/6
7.0 3/6
Stratified Random Sampling
• Sampling Distribution of Population Total
Stratum-I Stratum-II Stratum-I Stratum-II Stratum-I Stratum-II
1,2 4,7 1,8 4,7 2,8 4,7
4,7 4,7 4,7
4,7 4,7 4,7
7,7 7,7 7,7
7,7 7,7 7,7
7,7 7,7 7,7
Stratum-I Stratum-II Stratum-I Stratum-II Stratum-I Stratum-II
1,4 4,7 2,8 4,7 2,4 4,7
4,7 4,7 4,7
4,7 4,7 4,7
7,7 7,7 7,7
7,7 7,7 7,7
7,7 7,7 7,7
Stratified Random Sampling
Estimation of Population Mean and Population Total
• Results for Example 2.2 (SRS)
Population Total = t = 40
𝑡Ƹ = N 𝑦.
ത
𝑛 𝑆2
Var(𝑡)Ƹ = (1- )*
𝑁 𝑛
• Results for StrRS
𝑡Ƹ 𝑠𝑡𝑟 = σ𝐻
ℎ=1 Nh 𝑦
ത ℎ.
Var(𝑡)Ƹ = σ𝐻ℎ=1 N 2 V(𝑦
h ത ℎ)
Stratified Random Sampling
• Example 3.2
Population data is divided into four parts: yi = number of acres
devoted to farms in county i in 1992.
Stratified Random Sampling
• The data sampled from all four strata are in data file agstrat.dat. A
boxplot, showing the data for each stratum, is in Figure 3.1. Summary
statistics for each stratum are given below:
Data: agstrat.dat
Stratified Random Sampling (Example 3.3)
• Siniff and Skoog (1964) used stratified random sampling to estimate
the size of the Nelchina herd of Alaska caribou in February of 1962.
Stratified Random Sampling (Example 3.3)
Confidence intervals for stratified samples
• If either (1) the sample sizes within each stratum are large, or (2) the
sampling design has a large number of strata, an approximate
100(1−α)% confidence interval (CI) for the population mean 𝑦ത𝑈 is

• An approximate 95% CI for the total number of caribou is

54,497 ± 1.96(5840) = [43,051, 65,943].
Confidence Intervals in R
Stratified Random Sampling
Estimation of Population Proportion (SRS)
• Suppose we want to estimate the proportion of units in the
population that have some characteristic—call this proportion p.
Define yi to be 1 if the unit has the characteristic and to be 0 if the
unit does not have that characteristic. Then is estimated
by
σ𝑛
𝑖∈𝑆 𝑦𝑖
𝑝Ƹ =
𝑛
Stratified Random Sampling
• Sampling Weights in Stratified Sampling
1
wi = (for SRS)
𝜋𝑖
1
whj = (for StrRS)
𝜋ℎ𝑗

• For an SRS, the sampling weight for each observation is the same since all of
the inclusion probabilities πi are the same. In stratified sampling, however we
may have different inclusion probabilities in different strata so that the
weights may be unequal for some stratified sampling designs.
Stratified Random Sampling
• Sampling Weights

A stratified random sample from a population

with N = 500. The top row is Stratum 1; rows 2–
4 comprise Stratum 2; the bottom 21 rows are
Stratum 3. Units in the sample are shaded.
Stratum 1 has N1 = 20 and n1 = 10, so the
sampling weight for each unit in Stratum 1 is 2.
For Stratum 2, N2 = 60, n2 = 12, and the sampling
weight for each unit in Stratum 2 is 5. For
Stratum 3, N3 = 420, n3 = 20, and the sampling
weight for each unit in Stratum 3 is 21.
Stratified Random Sampling
• For the caribou survey in Example 3.3, the weights are:
ANY QUESTION???