non-parametric tests Flashcards
tests that do not require assumptions about distributions to be made are called:
non-parametric tests
non-parametric tests are not used as often because:
they do not have as much power as parametric tests, they are less precise, they can not be used to detect interaction effects.
having nominal/ordinal data, having heavily skewed data, having a small sample size, are all reason to use/not use non-parametric tests because?
use!, as these are all things that violate assumptions needed to do a parametric test
define these measure: nominal: Ordinal: interval: Ratio:
nominal: categories/classes, numbers have no real meaning
i.e: 1=male, 2=female
Ordinal: responses are ordered but intervals in between have no real meaning
e.i grades, 1st, 2nd 3rd (1st may of one by 1 point or 3, no distinction)
interval: scores are ordered with equal intervals, but there is no true 0
score on a depression test, difference between numbers is defined but scoring a 0 does not mean you have 0 depression it means you have the lowest amount.
Ratio: scores are ordered with equal intervals 0 exist as a real number
e.i: height, 0 would mean you were 0 cm tall etc. children; there is definable difference between 1 and 2 children and 0 means you have none
non-parametric test assumptions are:
each participant contributes only 1 data set (participates once)
data is ‘at least; ordinal i.e still fits into 1 of the 4 categories
distancee of distribution of data is approx the same
the non-parametric version of a t-test is a
mann-whitney U test or a wilcoxon rank sum test
the non-parametric version of a paired samples t-test is a :
wilcoxon signed rank test
the non-parametric version of a one-way between groups ANOVA is a:
Kruskal-williss one-way ANOVA
the non-parametric version of a one-way repeated measures ANOVA is a:
friedman test
