Sampling strategies and bias

Learning objectives

• Choose between random, systematic, stratified and time-based samples.
• Identify selection bias before it invalidates the study.
• Define a sample frame that matches the population of interest.

Sampling that represents the process

A sample estimates a population only when selection has no systematic bias. Stratify by style, size, colour, line, machine, operator, shift and lot when those factors matter.

A large convenient sample can be more misleading than a small honest one because size hides bias behind narrow confidence intervals.

Garment-factory example

Sampling only the top garments in a repaired bundle understates final defect risk. Stratifying by line × size × shift gives an honest picture of variation the buyer will actually see.

Method

1. Define the population and list the important strata.
2. Choose probability or rational sub-grouping.
3. Calculate a practical sample size for the target precision.
4. Log every missing observation and its reason.

Common mistakes

• Convenience sampling from the top of the pile.
• Assuming a large biased sample is reliable.