Non-Parametric Tests Flashcards
(28 cards)
Can non-parametric tests be described by equations?
No
They don’t depend upon the assumed probability function
Are parametric tests more powerful?
Yes
They can more readily identify group differences if they exist
When can you use parametric tests?
If the study design and the data has meet the prerequisites and assumptions
What are the prerequisites for parametric tests?
Dependent variable must be measured on an interval or ratio scale
Sample size sufficient (>20)
The population distribution from which the sample is drawn is known or assumed to be normal
The sampling distribution is known to be normal (don’t have to worry about this due to the central limit theorem)
Do non-parametric tests deal more with medians than means?
Yes, because the distributions of the sample do not have to be normal
Median gives us more information when data is skewed
What are the assumptions for parametric tests?
Distribution of the dependent variable in each of the sample group is approximately normal (confirm with a shapiro-wilk or kolomogorov-smirnov or histograms)
Variances (SD) of the dependent variable are not significantly different across samples (confirm with levene’s test)
How can you make non-parametric tests just as powerful as parametric ones?
More subjects
What is the non-parametric test equivalent for independent t-test?
Mann-Whitney U
What is the non-parametric test equivalent for dependent t-test?
Wilcoxon
What is the non-parametric test equivalent for ANOVA?
Kruskal-Wallis
What is the non-parametric test equivalent for RMANOVA?
Friedman
Is the Mann-Whitney U test a rank-sum test?
Yes
You combine all of the scores, rank order them, and test the difference between the sum (or mean) of ranks
What can Mann-Whitney U tests be used for?
To determine whether two independent samples are drawn from populations with the same median (not mean)
Are the other non-parametric tests sum rank?
Yes, except for chi-squared
How does the Wilcoxon work?
Rank the absolute values of differences in paired data (ignoring the positive and negative signs)
Assign the original valence to the corresponding rank
Sum the ranks with positive signs and negative signs
Use the sums of the positive and negative ranks to calculate a test statistic and determine the significance of the results
How does the kruskal-wallis test work?
If the kruskal-wallis (omnibus) test is significant, follow up with multiple pair-wise comparisons
Multiple Mann-Whiteney U’s with a Bonferroni adjustment
What does a p value of <0.05 on a shapiro-wilks or KS test mean?
Distribution is not normal
How does the Friedman test work?
If the Friedman (omnibus) test is significant, follow up with multiple pair-wise comparisons
Multiple wilcoxon tests with bonferroni adjustment
What does a p value of <0.05 on a Levene test mean?
Variances are significantly different between groups/intervals
Do all prerequisites and assumptions need to be met for a parametric test to be performed?
Yes
What is a chi-squared test?
Only used for nominal data
Most commonly used test in medical literature
Not optimal for ordinal data because it is underpowered - doesn’t take into account the order or magnitude of differences
How do you calculate expected frequency for chi-squared tests?
Sum the row and column counts and divide by the total number of cases
*what we would expect if it was completely random - no relationship between the two variables
What do you do after you have the expected values for each variable?
(Observed - expected)^2 / Expected
Calculate for each variable and add them together
How do you calculate degrees of freedom for a chi-squared test?
(rows-1)(columns-1)
*look up critical value on A-5 table on page 620
