Non-parametric tests Flashcards
A statistic pertaining to a population
‘Parameter’ e.g mean of population, SD of population, variance of a population
True or false - Non -parametric tests use frequencies and ranks not means and SDS
TRUE
When would you use a non parametric test? (IN EXAM)
- if the original scores may have unusually high variance
- an experiment occasionally produces infinite or undermine score e.g participant fails to solve a problem ( you can then rank the participants scores instead)
Name the Non - parametric tests that have parametric alternatives ( All ordinal data)
Wilcoxan Signed-Ranks test Mann Whitney U test Spearmans Correlation Kruskal-Wallis Test Freidman Test
Wilcoxan Signed-Ranks test
Concepts
Similar to repeated measure test, using ranks.
Requires the different scores to ranked from smallest to largest ( ignoring + -)
- uses signs and ranks
All the non-parametric procedures for ordinal data involve ranking scores from highest to lowest and then analysing their ranks
TRUE Kruskal- wallis wilcoxons rank sum mann-whitney friendman spearmans correlation
Non-parametric test that don’t have parametric alternatives. These require Nominal scaled data ( people in groups that cant be ranked)
Chi Square goodness of fit test chi square test of independence Point Biserial Correlation Phi Coefficient Binomial Test
Point bi-serial correlation use to measure?
the relationships between one variable (2 categories e.g gender) and a second with continuous score e.g height
Phi measures the strength of the correlation of variables in the chi square thus provides effect size. How many categories and variables does it test?
2 variables with 2 catgeories each e.g sex male and female and whether they watched, didnt watch the royal wedding
chi square test for independence
Is x related to y use formula, check if value is signifcant in table reject or not the null hypotheses
Cramers v test can be used to measure effect size
when chi square involves matrix larger than 2 x 2
Assumptions for the chi square. 1.-Independence of observations - people only counted in one category
2. ?
- size of expected frequencies must meet 5 or above. Should not use chi if this happens
The Binomial Test
Categories A and B and report as P and Q
- measure scale exist 2 categories
- only be in one of the two categories
- data is frequency of the individuals in each catgeory
The Binomial Test
Categories A and B and report as P and Q
- measure scale exist 2 categories
- only be in one of the two categories
- data is frequency of the individuals in each category
compare to z score
baby cliff experiment
