What do parametric tests aim to do?
Estimate sample means and variances and ask how likely it is the parameters are the same.
What do non-parametric tests aim to do?
Find out how likely it is that the data come from one distribution.
What do you use a non-parametric test?
When data is ordinal and a non-parametric test exists.
What kind of set of methods are non-parametric tests?
What kind of set of methods are parametric tests?
They use one framework: GLM.
What descriptive statistics are most appropriate with ordinal data?
A five number summary.
What does a sign test ask?
Is the sample median the same as a hypothesised value?
What are more powerful sign tests more likely to detect?
Differences that reflect true differences between populations.
What is Wilcoxon’s solution to finding probabilities with moderately sized samples?
Converting to a z-score.
What is different about a median in ordinal compared to scale data?
A, say, 2 unit difference below the median may not be the same as a 2 unit difference above the median.
What does scholastic ordering allow us to investigate?
Whether the data values in one group are likely to be smaller than the data values in a second group.
What do signed and Wilcoxin tests assume about samples?
That they are dependent.
What does the Mann-Whitney U assume about samples?
That they are independent.
When is it most important to use exact probabilities in the Mann-Whitney U?
When sample sizes are different.
What is the implication of Kolmogorov-Smirnov being sensitive to changes in means?
One distribution is shifted compared to the other, creating a large gap.
What is the implication of Kolmogorov-Smirnov being sensitive to changes in variance/range?
One distribution rises faster than the other, creating a large gap.
Which non-parametric test is functionally equivalent to 1-way between subjects ANOVA?
Which non-parametric test is functionally equivalent to 1-way within subjects ANOVA?
What is the main assumption of the Kruskal-Wallis?
That all distributions have the same shape.