Non-parametric alternatives to ANOVA

Question 1

non-parametric tests

Answer 1

• no population parameters
• no stringent assumptions
• can deal with small samples
• data can be measured on any scale

Question 2

parametric tests

Answer 2

• population parameters
• rigorous assumptions
• minimum sample size required
• require interval or ratio data

Question 3

example of non-parametric tests

Answer 3

chi square test
kruskal wallis test
Friedman test

Question 4

when to use non-parametric tests

Answer 4

• data is non-normal
• outliers
• DV isn’t interval or ratio
• assumptions are violated
• small sample size
• unequal sample sizes

Question 5

why not use all the time?

Answer 5

• parametric tests are more powerful and more likely to find differences
• means, SD and error variances vs rankings and frequencies

Question 6

Kruskal Wallis test is alternative to…

Answer 6

1 way between-subjects anova

Question 7

alternative to 1 way between-subjects anova

Answer 7

Kruskal wallis test

Question 8

what is the Kruskall wallis test/what does it do?

Answer 8

generalisation of mann whitney u

tests for sig. differences in 2+ groups of IV/factor on a continuous or ordinal DV

Question 9

SPSS outputs for Kruskall wallis test

Answer 9

descriptive stats (report median)
ranks (mean rank)
test stats (X^2(df)=KW value, p= )

Question 10

reporting results of non-parametric tests

Answer 10

• present medians in a table or text
• why was it selected (violated assumptions)
• test type, effect of IVs on DV
• test results (df, test statistic, p value)
• post hoc results

interpret

Question 11

follow up test for kruskall wallis test

Answer 11

• planned follow up test
• b-s, non-parametric alternative to independent t-test is mann whitney u
• apply bonferroni adjustment
• if KW test is sig. need to do followup pairwise analyses like in ANOVA

Question 12

non-parametric alternative to 1-way with-subjects ANOVA

Answer 12

friedman test

Question 13

what is the friedman test an alternative to?

Answer 13

non-parametric alternative to 1-way with-subjects ANOVA

Question 14

what is the Friedman test and what does it test for?

Answer 14

generalisation of wilcoxon signed ranks

test for sig. difference in 2+ groups of IV/factor on a continuous or ordinal DV

Question 15

SPSS outputs for friedman test

Answer 15

descriptive stats - medians
ranks - mean ranks = highest score
test stats - X^2(df)=chi-square, p value

Question 16

followup tests for friedman test

Answer 16

w-s/repeated measures = sign test (alternative to related pairs t-test)
bonferroni adjustment applied