What is Stratified Sampling & When is it Used?
What is Stratified Sampling?
Stratified sampling is a process used in market research that involves dividing the population of interest into smaller groups, called strata. Samples are then pulled from these strata, and analysis is performed to make inferences about the greater population of interest.
In order to fully understand stratified sampling, it’s important to be confident in your understanding of probability sampling, which leverages random sampling techniques to create a sample.
Stratified sampling is also commonly referred to as proportional sampling or quota sampling.
When is Stratified Sampling Used?
Stratified sampling is used when:
- A researcher’s target population of interest is significantly heterogeneous;
- A researcher wants to highlight specific subgroups within his or her population of interest;
- A researcher wants to observe the relationship(s) between two or more subgroups; and,
- A researcher’s goal is to create representative samples from even the smallest, most inaccessible subgroups of the population he or she is interested in.
When using stratified sampling, researchers have a higher statistical precision compared to when they elect to use simple random sampling alone. This is due to the fact that the variability within the subgroups is lower compared to the variations when dealing with the entire population at large.
Thanks to the statistical precision that stratified sampling provides, a smaller sample size is required, which can ultimately save researchers time, money, and effort.
How to Perform Stratified Sampling
The process for performing stratified sampling is as follows:
Step 1: Divide the population into smaller subgroups, or strata, based on the members’ shared attributes and characteristics.
Step 2: Take a random sample from each stratum in a number that is proportional to the size of the stratum.
Step 3: Pool the subsets of the strata together to form a random sample.
Step 4: Conduct your analysis.
An Example of Stratified Sampling in Context
Let’s imagine that a group of researchers is interested in distributing a survey to 50 students that are either juniors or seniors in traditional American high school.
The gender breakdown of the students is as follows:
Year | Boys | Girls |
Junior |
126 | 94 |
Senior | 77 | 85 |
Total | 203 | 179 |
The caveat with this study is that the researchers only have enough resources — time and money — to distribute their survey to a sample of 50 students.
Therefore, the sample must accurately represent the larger population of all students at the school in order for the results to be informative and actionable.
In this case, performing stratified sampling is the best option to deploy in this specific research study.
So…how many senior girls should be included in the 50 person sample?
By looking at the table above, it’s clear that if the researchers had access to the resources that would enable them to survey all 382 students at the high school, they would need to survey all 85 senior girls. But how does this translate to a sample of just 50 students?
To determine this, the researchers would simply take the fraction of 85/382 and multiply it by the sample size of 50.
This yields a result of 11.2, which when rounded down, represents 11 senior girls that need to be surveyed in order to have a representative sample.