W4 sampling and bivariate regression Flashcards
Universalistic questions
- “how does this work? what is the mechanism?”
- testing specific theories or hypotheses about the relationships between variables
- universalistic goals have been the usual goals of the studies you will have done in earlier courses
- Ex
Do impulsive adolescences drink more than less impulsive adolescences?
Particularistic questions
- “how many people think this? have this disorder? how many people would benefit from this treatment?”
- called descriptive hypotheses
- the primary goal is statistical estimation
- we try to capture the range of the underlying parameters (proportion of agreement, number of cases in population)
Ex
what is the prevalence of food addiction in Aus?
Sampling frame
- a list of people from the population of interest (aka our population elements)
- should ideally include the entire population and only those in the population
- traditionally this was the phone book, can be electoral role, students in 3003
Probability sampling
Based on random sampling approaches
- simple random sample (SRS)
- stratified sampling
- cluster sampling
Advantages of probability sampling
- we have a known or estimated probability of inclusion for each sample element
- the advantage here is that we know or can estimate the probability of inclusion then we can factor that into our final sample
- lower risk of sampling bias
- greater external validity
Disadvantages of probability sampling
- expensive
- often not feasible
Simple random sampling
- every member of the pop has an equal and known chance of being selected (referred to as pure random sampling)
- gold standard but requires very large samples and so can be very expensive and not always feasible
Ex
a lotto draw where all the numbered balls in the barrel represent sampling frames, each ball has an equal chance of being drawn
Stratified sampling
- in some stratified sampling the pop is divided in subpops in some meaningful way to ensure all subpops are sampled appropriately
- typically you create subgroups (aka strata) based on particular characteristics eg gender, profession, age group
- Using proportion of each subgroup you then sample the relevant number of p’s
Cluster sampling
- uses subgroups in the pop but rather than sampling individuals based on subgroup proportions, entire subgroups are sampled
Non-probability sampling
- convenience sampling
- purposive sampling
- snowball sampling
- quota sampling
Advantages to non-probability sampling
- cheaper
- easier to access
- much larger samples and therefore greater precision
Disadvantages of non-probability sampling
- unlike probability approaches, the probability of inclusion is unknown and therefore cannot be calculated
- very high risk of unknown bias
- unknown bias results in ambiguity of our results
- limits external validity (aka generalisability)
Convenience sampling
- involves recruiting a sample of p’s that the researcher has access
- friends, clients, from the clinic of a colleague, employees of the organisation
- most widely used type of sampling in psychology
- while this approach has unknown bias, we can increase precision with the much larger sample size that this approach affords
Purposive sampling
- also known as judgement sampling or known groups sampling
- the researcher makes a judgement based on their expertise to select the best sample
Snowball sampling
- involves having p’s recruit other p’s with an increasing number of p’s
Quota sampling
- effectively the same as stratified sampling however with non-random approaches
Bias and precision
- Bias is influenced by our sampling approach
- Precision is influenced by our sample size
- > And to an extent our sampling approach as well
We want to have samples that are:
- Low in Bias and
- High in Precision