Please enable JavaScript.
Coggle requires JavaScript to display documents.
quant W10 non-parametric tests - Coggle Diagram
quant W10 non-parametric tests
parametric inferential tests
assumptipons
underlying probability distributions (normal distributions)
DV measured at interval or ratio level- continuous data, numerical values
no outliers- can bias reuslts/means
homogeneity of variances (specific to independent t-test)- different people bring individual differences so variables around must be similar
linearity (specific to correlation)
what happens when assumptions arent met?
data might be non-normal (skewed/multimodal)
data not at required level
outliers
sample size too small
unequal sample size if using groups
Cohens guidelines
parametric test
Independent t-test
N>26 per group (to detect a large effect size), >64 (medium), >393 (small)
Related t-test
N>28 to detect a large effect size, >85 (medium), >783 (small)
Pearson's correlation
N>15 to detect a large effect size, >34 (medium), >199 (small)
look in advanced how many ppts needed to recruit, needed when looking for funding
specific sample size is needed- depends on effect size, larger effect size is easier to detect (bigger difference), Small/median effect size= more ppts needed as it'll be more difficult to detect
how do non-parametric tests work
distribution free- dont rely on normal distribution
ranks of data
outliers have little impact
They analyse the same research question/ hypotheses as parametric tests- only after data is collected a decision said to be made whether to use nonparametric or parametric
non-parametric tests
Mann-Whitney U
alternative to the independent samples t-test
Wilcoxon Signed Rank
alternative to the related samples t-test
Spearmans Rho
alternative to the pearsons product moment correlation