Stratified sampling is a sampling procedure in which the target population is separated into unique, homogeneous segments (strata), and then a simple random sample is selected from each segment (stratum). The selected samples from the various strata are combined into a single sample. This sampling procedure is sometimes referred to as “occasional fee sampling.” Read below about some of the considerations to remember for the best capture.
Stratified sampling is one of the types of probabilistic sampling that we can use. I invite you to continue reading to learn more about its weaknesses and strengths.
Selection steps for a stratified survey
There are eight main steps in selecting a stratified random sample:
- Define the target population.
- Identify the stratification variable(s) and determines the number of strata to use. The stratification variables must be related to the purpose of the study. If the purpose of the study is to make estimates of subgroups, the stratification variables must be connected to those subgroups. The availability of auxiliary information often determines the stratification variables that are used. More than one stratification variable may be used. Consider that as the number of stratification variables increases, the probability increases that some of the variables cancel the effects of other variables. In particular, no more than four to six stratification variables and no more than six strata of a variable should be used.
- Identify an existing test frame or develop one that includes information on the stratification variable(s) for each item in the target population. If the sample frame does not include information on the stratification variables, stratification would not be possible.
- Evaluate the sampling frame for undercover, overcover, multiple, and clustering, and make adjustments as necessary.
- Divide the sampling frame into strata, and categories of the variable(s) stratification, creating a sampling frame for each stratum. Within the stratum the differences should be minimized, and the differences between the strata should be maximized. The strata must not overlap, together they must constitute the entire population. The strata must be independent and exclusive of the subset of the population. Each element of the population must be in a single stratum.
- Assign a unique number to each item.
- Determine the sample size for each stratum. The numerical distribution of the items included in the sample across the various strata determines the type of test to implement. It can be a proportional stratified demonstration or one of several types of disproportionate stratified demonstrated.
- Randomly selects the specified number of items from each stratum. At least one element must be selected from each stratum to represent the sample; and at least two elements must be chosen from each stratum to calculate the margin of error of the estimates calculated from the collected data.
Proportional Stratified Sampling
There are two main subtypes of stratified sampling: proportional and disproportionate sampling. In proportionate stratification, the number of items assigned to various strata is proportional to the strata’s representation of the target population. That is, the sample size drawn from each stratum is proportional to the relative size of that stratum of the target population.
The sampling fraction is applied to each stratum, giving each population element an equal opportunity to be selected. The resulting sample is self-weighted. This sampling procedure is used when the research aims to estimate population parameters.
The researcher often wishes not only to estimate population parameters but also to do detailed analysis within a relatively small stratum and/or to compare strata with each other. Proportional stratified sampling may not result in some of the strata of this type of analysis.
Taking the example described in our table, it would not be possible to carry out a detailed analysis of the elements in zone 2 since only 12 of the elements are found in the sample. Furthermore, the comparison of the elements of zone 2 with the other zones would be doubtful.
Proportional stratified sampling is not a good choice of sampling to carry out this type of analysis. The disproportional may be a better choice.
Disproportionate stratified sampling
Disproportionate sampling is a procedure in which the number of elements included in the sample from each stratum is not proportional to their representation in the total population. The elements of the population do not have an equal chance of being included in the sample. The same sampling fraction does not apply to each stratum.
On the other hand, the strata have different sampling fractions, and as such, this sampling procedure is not an equiprobable selection. To estimate the population parameters, the population composition must compensate for the sample’s disproportion. However, for some research projects, disproportionate stratified sampling may be more appropriate than proportional.
Disproportionate sampling can be divided into three subtypes based on the purposes of our assignment. For example, it could be to facilitate analysis within strata, focus on optimizing cost, accuracy, or both accuracy and costs.
The objective of a study may require a researcher to carry out a detailed analysis of the sample strata. If proportional stratification is used, the sample size of a stratum is very small; therefore, it may be challenging to meet the study’s goals.
The proportional allocation may not produce a sufficient number of cases for this type of detailed analysis. One option is to oversample small or infrequent strata. Such oversampling would create a disproportionate distribution of sample strata compared to the population. However, there may be a sufficient number of cases to carry out the strata analysis required for the purposes of the study.
Strengths and weaknesses of stratified sampling
Stratified sampling has many of the strengths and weaknesses associated with most probability sampling procedures compared to non-probability sampling procedures.
Compared to simple random sampling, the strengths of stratified sampling include:
- Ability to estimate population parameters and make inferences within each stratum and comparisons between strata. Simple random sampling may not capture sufficient data on subgroups of interest. Stratified samples produce smaller random sampling errors than are obtained with a simple random sample of the same sample size. A stratified sample will result in a sample that is at least as precise as a simple random sample of the same sample size.
- Stratified samples tend to be more representative of a population because it ensures that elements from each stratum in the population are represented in the sample. Sampling can be stratified to ensure the sample is spread over geographic subareas and population subgroups.
- Using stratified sampling, the researcher’s knowledge about the population is taken advantage of.
- Using stratified sampling allows the researcher to use different sampling procedures within different strata.
What is the difference between stratified sampling, sampling, and quota sampling?
Stratified sampling and quota sampling are similar to each other. Both involve dividing the target population into categories and then selecting a certain number of items from each category. Both procedures have as their main objective the selection of a representative sample and/or the facilitation of subgroup analysis. However, there are important differences.
Stratified sampling uses simple random sampling. A sampling frame is needed for stratified sampling but not for quota sampling.
Advantages of quota sampling:
- It is the least expensive sampling method.
- It is widely used in polls and surveys by the media.
- The method assumes that the information we want to obtain is correlated with the population, but it is a hypothesis of representativeness that is difficult to prove.
In short, consider choosing stratified sampling if:
- It is possible to divide a population into two or more strata and build a homogeneous sampling frame for each stratum.
- Some subgroups of the population are very different from other subgroups.
- It is essential to minimize sampling error.
- The population is heterogeneous.
- A comparative analysis of the strata is desired.
Finally, we share an article on the characteristics of another type of sampling, systematic sampling.
Solve your research challenges with simple-to-create surveys that provide complex, real-time insights. Leverage custom point-and-click logic, advanced question types, and integrations and create mature, complex research models for choice-based research, consumer research, and more.
