choosing a statistical test Flashcards
what 3 factors are used to determine what statistical test is used?
- whether researcher is looking for a difference or correlation / association
- if a difference, what experimental design is being used
- level of measurement
decision 2: experimental design
- related designs: repeated measures, matched pairs
- unrelated: independent groups
what are the 3 levels of measurement?
- nominal
- ordinal
- interval
what is nominal data?
- data in the form of categories
- one item can only appear in one category (discrete)
what is ordinal data?
- distance between units of measurement are not equal
- lacks precision as it is based on subjective opinion rather than objective measures
how is ordinal data used in statistical testing?
raw scores are converted to ranks (ie. 1st, 2nd etc.) which are used in the calculation
what is interval data?
- data on numerical scales that include equal, precisely defined size
- most precise and sophisticated form of data in psychology
which statistical tests are parametric tests?
- related t-test
- unrelated t-test
- pearson’s r
why are parametric tests used?
- more powerful and robust than other tests
- can detect signficance in some data sets that non-parametric tests cannot
what 3 criteria must be met to use a parametric test?
- interval data
- data should be drawn from a population which would be expected to show a normal distribution for the variable being measured
- homogeneity of variance: sets of scores in each condition should have similar dispersion
how can you determine homogeneity of variance?
- compare standard deviations in each condition
- if they are similar, a parametric test may be used
- in a related design, it is general assumed that the 2 groups of scores have a similar spread
nominal: measure of central tendency and dispersion
- CT: mode
- D: n/a
ordinal: measure of central tendency and dispersion
- CT: median
- D: range
interval: measure of central tendency and dispersion
- CT: mean
- D: standard deviation
choose a test: nominal data, test of difference (unrelated design)
chi-squared