Sampling Methods and Cautions
What is Random? Random: Not Random: Drawing from a hat Rolling a die Using a number generator Not Random: “Picking” randomly First one to…(raise their hand, do a task, etc.)
Why Random? Everyone has an equal chance of being selected Helps to reduce bias by eliminating the human factor of selection Helps to account for the variability between people (even when we don’t know it’s there)
Types of Random Selection Simple Random Sample: Everyone gets a number and is selected via a random number generator
Types of Random Selection Stratified Random Sample: The population is broken up into strata (Male/Female, Freshmen/Sophomore/Junior/Senior) and a certain number of people are randomly selected from each strata.
Types of Random Selection Systematic Random Sample: A random number generator is selected to pick a person from 1-X. Each Xth person is then selected after that. For example, a die is used to select a number 1-6. Let’s say it comes up a 2. The second person would be selected for the experiment, and then the 8th, and then the 14th, continuing by 6.
Types of Random Selection Cluster Sampling: Choosing random clusters of subjects. Pickings pairs of Husband/Wife, Father/Dauther, entire families, neighborhoods, etc.
Types of Random Selection Convenience Sampling: “randomly,” or randomly choosing those conveniently close to survey. Surveyors at the mall, how we’ve surveyed so far (in the classroom), asking friends or neighbors
Population of Interest We always want to ask ourselves: Who is this generalizing to? If we take the heights of our class, are we trying to generalize to AP Stats students? To Juniors and Seniors at HSHS? To the school? To the county, the state, the country…? What is the population of interest?
Bias Bias is bad, yet unavoidable. We seek to reduce bias.
Types of Bias Selection Bias: When the researcher actively creates a bias in the way they select their subjects. Undercoverage Non-random
Types of Bias Response Bias Non-response Voluntary Response
Types of Bias Anything else that might throw off the data Using a placebo unnecessarily Not using a placebo when one should Not utilizing blind or double-blind methods when appropriate Etc.
Random Digit Table