Stratified Sampling

Stratified Sampling

Stratified Sampling

Description

It divides the population in homogeneous subpopulations or strata, depending on characteristics. Each member should be assigned exactly only one stratum.

Why to use

To obtain a sample that best represents the entire population

When to use

  • Imbalanced data
  • Reducing the overall variance
  • Ensuring diversity in the sample

When not to use

  • Unable to classify every member of the population in a unique strata

Prerequisites

  • Unique strata
  • Every member of the population should be assigned a strata

Input

  • Any dataset containing categorical variables.

Output

  • A sample best representing the population.

Statistical Methods Used

—

Limitations

  • Members are classified into multiple categories.