Week 8: Tests of Differences Non-parametric alternatives Flashcards
proper SHT for testing the mean difference between the two independent groups?
Student’s t-test
applied when the mean differences were compared by multiple factors over time?
Mixed ANOVA
an extension of a Student’s t-test for multiple groups?
ANOVA
proper SHT for testing the mean differences between the two related groups?
Paired t-test
an extension of a paired t-test for multiple points in time?
Repeated measures ANOVA
applied when the mean differences were compared by multiple factors?
Factorial ANOVA
applied when the mean differences were compared controlling for a confounding variable?
Analysis of covariance (ANCOVA)
Parametric data -
Data is called parametric if normally distributed with equal variances
Assumptions for data to be parametric: (3)
- Data level of measurement to be continuous
- Normally distributed
- Equal variances across groups to compare
What tests are the most powerful on parametric data?
Parametric tests (student’s t-test, ANOVA)
Name four non-parametric tests used when data is non-parametric:
- Mann-whitney U test
- Wilcoxon signed-rank test
- Kruskal-wallis H test
- Friedman test
Non-parametric tests work on what principle?
the principle of ranking the data from smallest to largest (1 assigned to smallest score and n to the highest score)
Mann-whitney U test used in place of what for non-parametric data?
- Student’s t-test
- test the mean difference between the two independent groups
Wilcoxon signed-rank test used in place of what for non-parametric data?
- paired t-test
- test the mean difference between the two matched groups
Kruskal-wallis H test used in place of what for non-parametric data?
- ANOVA
- test the mean difference between the three or more independent groups
Friedman test used in place of what for non-parametric data?
- repeated measures ANOVA
- test the mean difference between the three or more related groups
Non-parametric tests make what assumption on the distribution of data?
NO assumption
Non-parametric tests are (less/more) powerful than parametric test on parametric data
LESS
T/F Non-parametric tests are available when data is nonparametric at the cost of some power
True
Why might Non-parametric test may not be testing the null hypothesis of interest?
Null hypotheses that can be studied using non-parametric tests tend to be very restrictive
Non-parametric methods are focused on ____ rather than estimation.
Significance testing
T/F Non-parametric test can’t calculate the confidence interval
True
Mann-Whitney U test is used to compare (2, 3 or more) sets of scores that are (independent, related)
2
independent
The Kruskal-Wallis H test is used to compare (2, 3 or more) sets of scores that are (independent, related)
3 or more
independent
Wilcoxon signed-rank test is used to compare (2, 3 or more) sets of scores that are (independent, related)
2
related
Friedman’s test is used to compare (2, 3 or more) sets of scores that are (independent, related)
3 or more
related
Non-parametric test used to compare the difference between the mean alcohol consumption scores at the two institutions. What test used?
Mann-Whitney U test
Need a SHT for comparing means for three independent groups?
ANOVA or Kruskal-Wallis H test
Need a SHT for comparing means over time?
Repeated measures ANOVA or Friedman test
Need a SHT for comparing means for two independent groups?
Student’s t-test or Mann-Whitney U test