Ch 7 Flashcards
Generalizability
does the sample represent the population
External Validity is especially important in what type of claims?
frequency claims
Population
entire set of people or things/products of interest
Sample:
smaller set taken from the population to represent it
Census
sampling every member of the population
Population of interest
population to which we want our research findings to generalize
If a sample has good external validity, we can say
the sample “generalizes to” or “is representative of” the population
Biased sample (unrepresentative sample)
not all members of a population have an equal chance of being included
Unbiased sample (representative sample):
all members of a population have an equal chance of being included
Ways to get a biased sample:
- convenience sampling
- self-selection
Convenience sampling
sampling only those easy to contact
Self-selection
sampling only those who volunteer
Probability sampling (random sampling/ random selection)
all members of a population have an equal chance of being included
Nonprobability sampling
involves nonrandom, biased sampling
Simple random sampling
names in a pool and randomly selecting a predetermined number of names
Cons of simple random sampling
- difficult
- time consuming
- Impossible if you can’t enumerate population of interest
types of probability sampling
- simple random sampling
- Cluster sampling
- Multistage sampling
- Stratified random sampling
- Oversampling
- Systematic sampling
Cluster sampling
randomly choosing groups/ clusters within a population and using all people in the chosen groups
Multistage sampling:
2 random sampling stages:
- 1st randomly choosing groups/ clusters
- then choosing a random sample within each chosen group
Stratified random sampling
selecting demographic categories (strata) then randomly choose people within each category, proportionate to their % in the population
Oversampling
- selecting demographic categories (strata) then randomly choose people within each category, then over representing one or more groups
- have to adjust statistics on back end
Weighting
statistical technique to control for bias so that subgroups are properly represented according to their weight in the actual population
Systematic sampling
using a computer or random # table, choosing 2 random numbers (4 and 7 for ex) and then choosing the 4th person in the room and every 7th person until the desired sample number is reached
Random sampling (probability sampling)
creating a sample in a way that enhances external validity, so that each member of the population has an equal chance of being in the sample
Random assignment
used ONLY in experiments, to assign participants to groups at random (like treatment and comparison groups)
increases internal validity
Types of non-probability sampling:
- Convenience sampling
- Purposive sampling
- Snowball sampling
- Quota sampling
Convenience Sampling
samples that are easy to access
- common in behavioral research
Purposive sampling
using only the type of people the researchers want to study, in a non-random way
Snowball sampling
variation of purposive sampling in which participants are asked to recommend other participants
Quota Sampling
similar to stratified random sampling-
- identifies subsets of interest, then set a target # for each subset. Sample non-randomly until quotas are filled.
when a non-probability sample is used, ask:
Are the characteristics that make the sample biased relevant to what’s being measured?
if a sample isn’t externally valid, it…
has unknown external validity
ways of saying a sample is random or not
- Unbiased sample/ biased sample
- Probability sample/ nonprobability sample
- Random sample/ nonrandom sample
- Representative sample/ unrepresentative sample
are large samples more representative than smaller samples?
- Large samples are not more representative than smaller samples (If the larger sample is not more representative than a smaller sample would be)
- A researcher chooses a sample size for a poll in order to optimize the margin of error for the estimate (the degree of sampling error)
