Chapter 4: Collecting Data

Multiple sampling techniques are presented with specific applications: simple random sampling assigns equal selection probability to every unit, stratified random sampling divides the population into homogeneous subgroups before sampling within each stratum to improve precision, and cluster sampling groups nearby or related units together for practical efficiency when populations are geographically dispersed or large. The chapter identifies critical sources of bias that compromise study validity, including undercoverage when certain population members lack selection opportunity, nonresponse bias when selected participants decline participation, and response bias arising from question wording, interviewer effects, or social desirability influences. Moving to experimental studies, four fundamental design principles emerge: establishing comparison groups to measure treatment effects, employing random assignment of participants to treatments to balance confounding variables, implementing control conditions as baselines for effect measurement, and replicating studies to ensure results are not due to chance variation. Specific experimental designs accommodate different research contexts—completely randomized designs assign all participants randomly to treatment groups, randomized block designs first organize participants into similar blocks before randomizing within blocks to reduce variability from known factors, and matched pairs designs pair similar participants and assign one to treatment and one to control. The chapter explains how blinding (both single-blind and double-blind) and placebo controls neutralize psychological influences and investigator bias. Finally, the distinction between observational studies that measure associations and experiments that establish causal relationships receives careful attention, along with ethical obligations in research design and the appropriate scope of statistical inference when generalizing from samples to populations.