Non-Parametric Tests (and parametric parts) Flashcards
When would you use a parametric test?
What’s the criteria?
(4 parts)
- You have normal distribution for your data
- Variances of data populations are equal
- Interval level of data
- No extreme scores (small range)
Non-parametric tests
What are the 5 tests?
- Mann-Whitney U Test
- Wilcoxon Signed Ranks Test
- Chi-square
- Binomial Sign Test
- Spearman’s Rho
Non-parametric Test
What 3 questions should you ask to help decide the test you need to do?
- What level of data do you have?- nominal or not
- Testing for a difference or correlation?- may say in question
- Experimental design?- Independent measures or repeated measures/ same or different participants
Non-parametric tests
What phrase can be used to remember the criteria?
- L= Level of data
- ION= Interval, Ordinal, Nominal
- D/C= Difference or correlation
- ED= Experimental Design
Mann Whitney U
What criteria is required for this test?
- Not nominal data
- Different participants- Independent measures
- Difference
Mann Whitney U
What are the steps to calculate?
- Rank scores from lowest to highest
- Use formula to find U
- Calculate the smaller U value
- (Given formula sheet)
Mann Whitney U
What is the level of significance?
- U value is less that critical value= reject null hypothesis as it’s significant
- U value is higher than critical value= accept null hypothesis as it’s insignificant
Wilcoxon Rank
What is the criteria for this test?
- Not nominal
- Repeated measures design (same participants)
- Difference
Wilcoxon
What are the steps to calculate?
- Calculate the difference between the values across the different conditions
- Rank differences
- Sum up the positive and negative ranks
- T is the smallest of the values
Wilcoxon
What is the level of significance?
- If W value is less than the CV= reject null hypothesis (it’s significant)
- If W value is more than the CV= accept null hypothesis (it’s not significant)
Chi-square
What is the criteria for this test?
- Nominal data
- Independent measures (different participants)
- Difference
Chi-square
What are the steps to calculate?
- Add totals for each column
- Calculate observed and expected frequencies (E= RxC/T) for each individual observed values
- Input each EV into formula then add values to get x2
- Find the df at 0.05 and check table of significance
5.
Chi-square
What is the level of significance?
- If calculated chi is less than CV= accept null hypothesis (insignificant)
- If calculated chi value is more than CV= reject null hypothesis (significant)
What is the criteria for this test?
(Binominal sign)
Nominal data
Difference
Same participants (repeated measures)
What are the steps to complete? (binominal sign?
- Calculate direction of difference between conditions (e.g. put a minus sign - after values in column 1 and keep consistent)
- Identify which sign is the least common (S+ S-)
- Look at CV table to obtain critical value for S
Binomial Sign
When is the result significant?
If S is less than or equal to the CV= result is significant
What is the criteria to calculate? (Spearman’s Rho)
Not nominal
Correlation
What are the steps to calculate? (Spearman’s Rho)
- Rank scores (each condition separately)
- Find difference between the ranks
- Square the differences (D)
- Add all of squared differences (D2)
- Use formula to calculate Rs
- Look at CV value table to compare Rs to CV
Spearman’s Rho
When is the result significant?
If Rs is larger than or equal to the CV the result is significant