Chapter 7 Flashcards
Let’s say I want to survey Canadian university
and college students on attitudes towards plagiarism and cheating.
I could survey this entire population (i.e., do a
census ), but that doesn’t sound very feasible. If I ask a sample of this population, it’s important that I acquire a representative (unbiased)
How to end up with an unrepresentative sample.
1) One route is to employ convenience sampling , using those who are easy to contact and available to participate. This is extremely common within psychology research, especially with association and causal claims, and that is an accepted limitation of the research.
2) Another weakness of online polling is
self selection : the sample is composed of those who chose to participate. Are those people different than those who did not participate?
A representative sample can be acquired using
probability
probability sampling (random sampling), a collection of different techniques that all have the same goal: a representative sample.
Methods of probability sampling
1) One is random sampling: each possible participant is assigned a number; then, random numbers are chosen using a random number table or a random number generator.
2)The talking study used systematic sampling , with a random starting point and a fixed interval (e.g., “every 100 th student on the list”). Very similar to random sampling, but with a numerical rule. So you might again assign every student in your population a number.
Instead of using a random number generator, you might start with student #25 and then select every 100 th student going forward.
3) In cluster sampling , clusters of participants within the population are randomly selected, and then everyone within each selected cluster is surveyed.
4) Multistage sampling is similar: clusters are randomly selected, and then participants within those clusters are randomly sampled. A random sample within each random sample.
5)Stratified random sampling accomplishes this by setting goals related to separate categories (strata) within the sample. E.g., science majors (20%)
and everyone else. Otherwise, it still uses random sampling.
6) Maybe you’re particularly interested in the opinions of sociology students, but they only make up 2% of students. Oversampling works just like the last one,
but your goal is higher than the actual percentage (say, 20%) … and, again, you would still use random sampling.
cluster sampling
Two levels:
- Which universities? Random selection
- Which students? All of them
Multistage sampling:
Two levels:
1. Which universities? Random selection
2. Which students at those universities? Random selection
Also an example of multistage sampling:
Three levels:
1. Which universities? Random selection
2.Which majors at those universities? Random selection
3.Which students within those majors at those universities?Random selection
Comparison of stratified sampling and oversampling
stratified: The real % and the target % are the same! oversampling: The target % is higher than the real %!
probability sampling
1) Simple random: Random selection among
every possible participant.
2) Systematic: Similar but with some kind
of rule (e.g., every 50th)
3) Cluster, multivariate: Both use clusters: within a
cluster, it’s either everyone or random selection.
4) stratified random: Making sure a certain
group is accurately represented
5) oversample: Similar, but making sure
a certain group is overrepresented.
Specific types of nonprobability sampling (some types of people are systemically left out)
1) convenience
2) quota sample: The non random equivalent to
stratified random sampling, aiming for a certain % goal.
3) purposive sample: Recruiting a certain group of interest via non random means.
4) snowball sample : Purposive with the addition of having participants help recruit via word of mouth.
reality is that a representative sample is a
good goal for frequency claims, but less of a concern for
good goal for frequency claims, but less of a concern for
association and causal claims.
