STAT 100 Lecture Notes - Lecture 2: Response Bias, Selection Bias
Document Summary
Probability sampling: simple random sampling, not a guaranteed representation, stratified random sampling, stratify people into groups based on things such as sex and race, take a random sample from each strata b. i. To represent each group: cluster sample, sampling taken as a random cluster a. i. Row 1 represents cluster of rows 1-4 a. i. 1. Representing more than one at a time m&m"s. Volunteer, picking people out of the hub: does not represent population. Results allow the researcher to extend those results (infer) back to the population. In random sampling we can also account for sampling error: margin of error. The approximate margin of error is 1/square root of n where n=sample size. Ex: 100 people, 1/square root of 100, 1/10, 10% margin of error. A common target margin of error for a sample survey is 3% (+-3%) You need about 1,000 people in your study: you are going to need to survey more because some may not answer.
