Unit 3 of 12
Collecting Trustworthy Data
- 4 lessons
- Videos 13 to 16
How data are collected decides which conclusions they can support. This unit covers sampling bias, random sampling methods, observational studies and randomized experiments.
A larger sample does not remove bias, so the unit begins with undercoverage, nonresponse, voluntary response and measurement bias. Simple random, stratified, cluster and systematic sampling then use chance in different ways. In an observational study an association can come from confounding or reverse causation. Random assignment makes treatment groups comparable on average over the possible assignments and supports a causal comparison, while random sampling supports generalizing to the sampled population. Blocking and matched pairs can make the comparison more precise.
Every lesson will be one video. The videos are coming soon, and each lesson page already has the definitions and an example for every key term.
Sampling Bias and Random Sampling
Sampling Bias and Representativeness
How a sample can systematically miss its target population through the sampling frame, undercoverage, nonresponse, voluntary response or measurement, and why a larger biased sample remains biased.
Video 13
Random Sampling Methods
How simple random, stratified, cluster and systematic sampling use chance, how inclusion probabilities lead to survey weights, and how the design effect compares a design's precision with simple random sampling.
Video 14
Observational Studies and Experiments
Observational Studies, Confounding and Causation
How an observational study differs from an experiment, what explanatory and response variables and a causal effect are, and how confounding, reverse causation and selection can make an association differ from the causal effect.
Video 15
Randomized Experiments, Blocking and Matched Pairs
How random assignment of experimental units, a control group, placebos and blinding support a causal comparison, how blocking and matched pairs can make it more precise, and how random assignment differs from random sampling.
Video 16