100 likes | 285 Views
Sections 1.4: Other Effective Sampling Methods 1.5: Bias in Sampling 1.6: The Design of Experiments General goals Collect data effectively Avoid bad data Clever ways to best answer the question at hand.
E N D
Sections • 1.4: Other Effective Sampling Methods • 1.5: Bias in Sampling • 1.6: The Design of Experiments • General goals • Collect data effectively • Avoid bad data • Clever ways to best answer the question at hand
A local news radio station invites morning listeners to respond to a opinion question about a current news issue. The listeners are given one of two phone numbers: one for if you “agree” and another if you “disagree”. This is an example of: • Stratified sampling • Systematic sampling • Cluster sampling • Convenience sampling
Suppose I wanted to find out HSU students’ opinion about their knowledge of the Jack Pass for the local bus system. Using student records, I randomly selected 20 first year students, 20 second year students, 20 third year students, and 20 student who were fourth or greater year students. This is an example of: • Stratified sampling • Systematic sampling • Cluster sampling • Convenience sampling
What would a possible cluster sampling design look like for surveying students about their bus use?
In attempt to predict the 1936 presidential election, The Literary Digest surveyed a sample collected from its readers and from a registry of car owners and telephone users. The survey predicted the Republican candidate Alf Landon would easily beat Democrat president Franklin Roosevelt. (source: Wikipedia). You’ve likely never had heard of Alf Landon, so obviously the survey was off in its prediction. What went wrong?
Where did it go wrong? • Nonresponse bias: Only 2.3 million of the 10 million readers responded to survey • Sampling bias: The ownership of cars and phones excluded many poor and/or rural people.
Suppose a public health survey is being designed to learn about people’s sexual behavior and their perceived risk for acquiring and/or transmitting STDs. Question: What type of bias(es) should researchers be concerned about? • Sampling & Nonresponse • Nonresponse & Response • Sampling & Response • Sampling & Nonresponse & Response
To test the effect of caffeine and blood pressure (bp), 100 volunteers were given a placebo pill one day and then had their bp measured. Likewise, on an adjacent day each person was given a caffeine pill and their bp measured. The order of placebo or caffeine was randomly set.
This is an example of: • Matched-pairs design • Multistage sampling • Double blind study • Systematic sampling
Some big ideas about designing a study: • Determine the question. • Determine the parameter of interest. • Determine the population. • Be aware of and avoid bias • Control for what you can (blocking, matched-pairs, etc) and randomize the rest. • Think of confounding variables, especially in observational studies. • Play “Devil’s advocate” and then redesign.