Ch 7 Lecture Flashcards
Random sampling
Drawing a sample using some random method; ensures members have an equal chance of being in the sample; good for external validity
Random assignment
Using random method to assign participants into different groups; ensures groups have same kinds of people in them; good for internal validity
External Validity vs Unknown EV
- EV: unbiased, probability, random, representative sample
- UEV: Biased, nonprob, nonrandom, unrep sample
Cluster sampling
- Randomly selects clusters of participants from already existing groups
- Collects data from everyone in the selected clusters
- 1 random sample (clusters)
Multistage Sampling
- Same as cluster, except instead of using every participant in randomly chosen clusters, randomly sample people w/in each cluster
- 2 random samples (clusters and people w/in them)
Quota Sampling
- Identifies categories in the ppn
- Sets a target # for each category
- Nonrandomly selects indivs w/in each of the categories
Stratified Random Sampling
Same as quota, except you randomly select indivs w/in each of the categories
Cluster vs SRS
With cluster, you collect data from everyone in a few categories, whereas with SRS, you collect data from a few people from every categories
When is randomness crucial
Frequency Claims (still important for the other two but not as much of a priority, as often relationships are similar to one another & other tests will be done later to determine generalizability better)
Are bigger samples better?
- The size of a sample is much less important for EV than how that sample is selected (randomly or not)
- What it helps is statistical validity, as a greater sample makes the margin of error smaller
Margin of Error
The degree of sampling error in a study’s results
Sufficient Sample Size
1,000 is sufficient as it has a margin of error of plus or minus 3%; can even be 2,000 maximum with a MofE of plus or minus 2%
