### What is Stratified Sampling?

In Stratified Sampling, researchers divide a sample based on specific characteristics, such as race, gender, and location, into homogeneous subpopulations, called *strata*. Each member of the studied population should belong to exactly one stratum.

The researchers then sample each stratum using another Probability Sampling method such as Cluster sampling or Simple Random Sampling. This allows them to estimate statistical measures of each subpopulation.

This is important for generalizability, validity, and avoiding biases in research such as under coverage.

### How to Use Stratified Sampling

You must be able to divide your population into mutually exclusive subgroups. This means that every member of the sample can be classified into **only one** group.

When you think that different subgroups may have different average values, stratified sampling is the most appropriate method. It can have several advantages.

- Ensure that your sample is diverse

This ensures that the sample reflects your population’s diversity. This is possible, but unlikely to happen with other sampling methods like simple random samples.

- Ensure similar variance

You need to use the same sample size in each subgroup if you want your data from each group to be similar.

You might get a small sample size with other sampling methods because certain subgroups are less common.

- Reduce the variance of the population

Even though your population as a whole may be heterogeneous in nature, certain subgroups can be homogeneous.

If you study how a new educational program affects test scores, it is likely that both the original scores of the children and any changes in their scores are highly correlated to family income. Scores are often grouped according to family income.

In this situation, a stratified sample allows you to measure variables more precisely, and with a lower variance for each subgroup, as well as for the entire population.

- Variety of data collection methods

You may have to use different methods when collecting data from different subgroups.

You may choose to do a study on rural subjects via mail, while you sample the urban subjects door-to-door.