Lecture 9 Flashcards
What are the 2 types of non-parametric correlation methods?
- Spearmans rank
- Kendalls tau
What is the method of a spearmans rank correlation?
Step 1: Rank transformation of x and y
Step 2: Pearsons r correlation on the transformed variables
Step 3: Produces a correlation coefficent: P
What data (normal/ non-normal) can spearmans be used for?
Both
What type of relationship does spearmans measure?
Monotonic
What data (normal/ non-normal) can Pearsons be used for?
Just Normal
What type of relationship does Pearsons measure?
Linear relationships
What is the method for Kendalls Tau
Step 1: Rank transform variables
Step 2: count concordant and discordant pairs
Step 3: Correlation coefficient denoted by greek letter tau (T)
What does the Tau-A correction do to ties?
No adjustment for ties, just keeps it in range -1 to +1
What does the Tau-B correction do to ties?
Adjusts ties to keep range -1 to +1. The most common. Easy to interpret
Do Spearmans and Kendalls yield similar results?
Yes typically similar results
what are the advantages of Kendall’s Tau?
√ More robust for significance testing in smaller sampels
√ Particularly suitable for discrete + Ordinal data
Evaluate ranked group difference tests
√ Ranking method provides robustness against outliers
√ Doesnt require normally distributed data
√ Can use with ordinal, unlike t -test
X - has less power than t tests
What are the 3 non-parametric tests for group differences?
Binomial test, Mann-Whitney u-test, Wilcoxon signed-rank test
When would you use binomial test?
With binomial data
How to you calculate correlation tests in SPSS?
Analyse - correlate - Bivariate - add variables - select method (Pearson, spearmans, kendalls)
What is the method for wilcoxon?
Step 1: Work out the difference between a pairt and allocate it a sign (+/-)
Step 2: rank them by differnce and multiply rank with sign
Step 3: Sum of signed ranks gives the value ‘W’
Step 4: The W and DF and used to derive the ‘Z statistic’ and significance ‘p’
- SPSS does all this
When would you use Wilcoxon?
Repeated measures design (its an alternative to the related t-test)
When would you use Mann Whitney?
For indepedent groups
What is the method for Mann whitney
Step 1: Combine all samples and rank all values
Step 2: Add up ranks which came from each sample, to find sum of ranks for each sample
Step 3: This gives you a ‘u’ value
Step 4: u and df used to derive z, which is used to get p
What can the ‘u’ value from the mann whitney test range from?
0 (complete seperation between groups)
to
n1 X n2 (no seperation between them - accept null)
How do you calculate Wilcoxon in SPSS?
Analyse - non parametric tests - legacy dialogues - 2 related samples - enter variables - select wilcoxon
How do you calculate Mann whitney in SPSS?
Analyse - non parametric tests - legacy dialogues - 2 independent samples - enter variables - select mann whitney