Variables, sampling, reliability and validity

simple random sampling

every participant has an equal chance of selection

systematic random sampling

randomly select first person, divide population by size of sample and create interval to select sample

stratified sampling

divide pop to subpop (strata) then randomly sample from there

makes sample profile match actual ops and includes small subpop

multistage cluster sampling

larger sample chosen first and sub sample chosen for study

non probability sampling looks at

specific populations under study

quota sample

non probability equivalent of stratified random sample

purposive sampling

based on knowledge of the population/study purpose

snowball

hard to study populations, collect data from members and ask for other contacts

convenience

easiest method, students enrolled in specific course

Larger sample sizes are needed when

heterogenous, needing multiple categories, want narrow confidence interval

6 rules for sampling

if under 100, use whole population, larger is better, do literature review, use power table, expect small effect, use sample size calculator like g power

test reliability

stability over time (>.70)

reliability

consistency/repeatability of a measure

split half reliability

divide test to 2 halves and compare results - internal consistency of a measure (>.70)

cronbach’s alpha (inter item reliability)

how well different items on test measure same construct

0-1.00

interrater/observer reliability

consistency across raters (>.90)

validity

are we measuring what we think we are?

face validity

on the face of it, how does my research relate to the construct?

content validity

extent measure represents balanced sampling of relevant dimensions

criterion related validity

checks performance of measure against external criterion - concurrent and predictive

construct validity

measure relates to theoretical construct of interest

convergent - to similar constructs

divergent - unrelated to different measures

separate

maximise variation between groups/level of IV

compress

minimise variation within groups/level of IV

error variance (variability)

3 sources: measurement error, individual differences, other factors in environment

nuisance variables

noise creating, affect DV, reduces power through creating bias

confounding variables

affects both IV and DV, reduces internal validity, try to control by eliminating, keeping constant or building into study

parametric manipulation

varying specific v across range of values to observe effects