# Stratified sampling: An effective approach for dividing the population into subgroups

In research design, the practical formulation is essential and hence must be done with immense care. But during the formulation of study on practical grounds, a large number of problems pertaining to determining the characteristics of the whole population or the part of the population creeps in accidentally. To encounter such problems, a technique known as sampling is employed in research.

Simply put, the sampling technique is a process of drawing a definite number of observations or individuals from the total population under investigation. Sampling methods are of various types, for example, snowball sampling, random sampling, etc. which can be used for different purposes. However, if the individuals in the population need to be defined using specific criteria, then we generally rely on the stratified sampling approach.

Stratified sampling involves the segregation of population into smaller subsets known as strata. In this type of sampling method, the strata are formed based on the characteristics or attributes of the individuals. Stratified sampling is widely used under the following circumstances.

- Specific subgroups within the population of interest need to be highlighted
- The target population is significantly heterogeneous
- A relationship between two or more subgroups need to be determined
- Representative samples from inaccessible subgroups must be created

Stratified sampling requires higher statistical precision when compared to any other form of sampling method. This is because the variability within the subgroups is lesser the variations of the whole population.

Stratified sampling is of two types.

**1. Proportionate stratified sampling** – In this type of sampling method, each stratum sample size is directly proportional to the population size of the whole population of strata. This means that each strata sampling includes similar sampling factor. The proportionate stratified sampling is represented by an equation,

n_{h} = ( N_{h} / N ) * n,

Where,

n_{h }is the sample size of ‘h^{th}’^{ }stratum

N_{h }is the population size of ‘h^{th}’^{ }stratum

N is the size of whole population

n is the size of entire sample

For example. If you have 3 strata and the sampling fraction is ½, then the final sampling sizes are as given in the table.

Sampling | A | B | C |

Population size | 200 | 300 | 400 |

Sampling fraction | 1/2 | 1/2 | 1/2 |

Final sampling size | 100 | 150 | 200 |

Here, irrespective of sampling size the sampling fraction will remain the same.

**2. Disproportionate stratified sampling** – Typically, the sampling fraction is a differentiating factor between proportionate and disproportionate stratified sampling. Unlike the proportionate method, this type of sampling approach has different sampling fraction.

Consider an example where you have three sample sizes and different sampling fraction. Then the final sampling size is as follows.

Sample | A | B | C |

Population size | 200 | 500 | 1000 |

Sampling fraction | 1/4 | 1/2 | 1/4 |

Final sampling size | 50 | 250 | 250 |

In this type of sampling, if the fractions alloted are not precise, then the results may be biased due to under or over presented data.

Stratified sampling technique works best when the population is a finite number, includes subgroups and the strata within the population is non-overlapping.

The steps included in this sampling technique are:

**a) Describe target audience** – The first step in the define the target audience and decide the sample size.

**b) Recognize stratification variables** – This is followed by recognizing the stratification variables and determine the number of strata to be utilized. The stratification variables must align with the major objective of the research. If there is any additional information in the study, then such information to plays a role in determining the stratification variable.

For example, if the objective of the study is to understand the subgroups, then the variables will relate to the subgroups and the information pertaining to the subgroups will influence the variables. Ideally, a study should include only a maximum of 6 strata and 4-6 stratification variables.

**c) Develop sampling frame** – Next, create a new frame or use existing sampling frame that includes all the necessary information of the stratification variable for the elements in the target audience.

**d) Evaluate the sampling frame** – Assess the sampling frame on the basis of grouping, overlapping and make the required changes.

**e) Assign unique numbers to element of strata** – Ensure that each stratum is unique, and the difference between the two strata is different from each other. Then, assign a unique and random number to each element.

**f) Identify the strata size** – Determine the study requirements and figure the size of the stratum. The numerical distribution will identify the type of sampling. Next, choose random samples and form the required sample.

Note: every element in the population must fit into one stratum. Put other words, an individual cannot be in more than 1 group.

Today, stratified sampling is popular among the research scholars because of the number of benefits such as improved accuracy, exactness nature of sampling technique, and ability its to cover the maximum population.