PSYC 202 Lecture 2: WEEK 3
15 Jun 2018
School
Department
Course
Professor
WEEK 3
Ideal sampling process:
1. Units have known and non-zero probability of being included in your sample
2. Unbiased
3. Independent
a. The selection of one sampling unit can’t influence the selection of others
4. Each sample has equal chance of being selected
Observational studies:
- Simple random
- Stratified
o First identifying strata and then selecting units from the strata
o Strata is not random – identified ahead of time
o E.g. selecting people from provinces to find out income
▪ Sampling unit is person
- Clustered
o Randomly selected clusters and then select units from the clusters
o E.g. randomly select cities in Canada and then selecting people from the selected
cities to find out the income
▪ Sampling unit is city
- Case-control
o Retrospective design
o Want the control to be as similar to the thing you’re studying
- Cohort
o Following a group over years
o More expensive
o Prospective design
- Prospective vs. retrospective
o Retro--
- Correlation
Experimental studies:
- Control treatments
- Blocking
- Blocking multiple factors
- Pseudo replication
Single factor treatment: measuring and focusing on one factor. Only for experimental studies.
Experimental – units are randomly assigned to treatments
Document Summary
Ideal sampling process: units have known and non-zero probability of being included in your sample, unbiased, independent, the selection of one sampling unit can"t influence the selection of others, each sample has equal chance of being selected. Stratified: first identifying strata and then selecting units from the strata, strata is not random identified ahead of time, e. g. selecting people from provinces to find out income, sampling unit is person. Clustered: randomly selected clusters and then select units from the clusters, e. g. randomly select cities in canada and then selecting people from the selected cities to find out the income, sampling unit is city. Case-control: retrospective design, want the control to be as similar to the thing you"re studying. Cohort: following a group over years, more expensive, prospective design. Single factor treatment: measuring and focusing on one factor.
