PSYC 202 Chapter Notes - Chapter 3: Statistical Population, Sampling Error, Confounding
15 Jun 2018
School
Department
Course
Professor
WEEK 3 VIDEO NOTES
Sampling and sampling error
Ideal sampling process
- Units have known and non-zero probability of being included in your sample
- Unbiased
- Independent
o Selecting one unit to be in the sample shouldn’t affect whether other people are in
the sample
- Each possible sample has equal chance of being selected
o Making sure individuals are mixed properly
Selection of units: every unit in the statistical population must have a chance of being in the
sample
Bias: selection of units cannot inadvertently favour one outcome over another on average
Independence: selection of one unit cannot influence the probability that another unit is selected
Equal chance for all samples: every combination of units must be possible in the sample
Sampling error
- Natural variation from one random sample to another
- Not a mistake
Designing observational studies
Observational studies
- Characterize an existing population
- Correlative, but not causal
- Confounding variables
o Variables you haven’t observed but are likely driving the relationship between
variables that you have observed
- Retrospective vs. prospective
o Retrospective – looking back
▪ Prone to confounding errors
▪
o Prospective – looking forward
Designs for observational studies
- Simple random survey
Document Summary
Units have known and non-zero probability of being included in your sample. Independent: selecting one unit to be in the sample shouldn"t affect whether other people are in the sample. Each possible sample has equal chance of being selected: making sure individuals are mixed properly. Selection of units: every unit in the statistical population must have a chance of being in the sample. Bias: selection of units cannot inadvertently favour one outcome over another on average. Independence: selection of one unit cannot influence the probability that another unit is selected. Equal chance for all samples: every combination of units must be possible in the sample. Natural variation from one random sample to another. Correlative, but not causal: variables you haven"t observed but are likely driving the relationship between variables that you have observed. Retrospective vs. prospective: retrospective looking back, prone to confounding errors, prospective looking forward. Simple random survey: every unit is selected at random from the statistical population.
