Flashcards
Non-parametric tests
Cards in this deck (21)
- 1What are assumptions?Assumptions cover criteria which must be met for a parametric test to be accurate.
- 2There are 4 main assumptions that must be met for parametric tests:- Normally distributed data
- Homogeneity of Variance
- Interval Data
- Independence
Normally distributed = symmetrical about the mean - 3What are nice data and not nice data?- Nice data; Normally distributed
Normal analysis procedure
- Not nice data;Not normally distributed
Separate analysis procedure - 4When do you use a non-parametric tests?- In certain situations we can’t correct problems in data…
- This is when we require non parametric test
- “Assumption free tests”
-These test generally rely on ranking the data E.g. lowest score = 1, next highest score = 2 etc… - 5So what is a non-parametric test?- A family of statistical procedures which do not rely on the restrictive assumptions of parametric tests.
- IN PARTICULAR – They do not assume that the sampling distribution is normally distributed - 6Is power an issue?Statistical power = the ability of the a test to find an effect that genuinely exists
- 7What does ranking the data mean?- Ranking the data means we lose information relating to the magnitude of differences
- 8What is the first theory?- Difference between two groups
-Equivalent of Independent T test
- These tests have different groups
- The ranking system works across all the data from both group
- The ranks are then added up in order to assess if there is a difference between groups - 9What is the non-parametric equivalent of the independent samples t-test?- Mann Whitney
- 10What is the Mann Whitney test?Assesses the difference between two independent samples
Tests whether the populations from which two samples are drawn have the same location - 11How do you conduct a Mann Whitney u test?LOOK ON NOW LECTURES (MEDIA) AND LEARN
- 12What is the second theory?- Wilcoxon rank test is based on the differences between scores in the two conditions you’re comparing
- Once the differences have been calculated, they are ranked and sign of difference (positive or negative) is assigned to the rank
- If the score is zero then the scores are the same for the two conditions and the data can be excluded
- The ranks are then collected from positive differences to get the sum of positive ranks
- We the do the same for negative ranks…
These are used to calculate the test statistic - 13What is the Wilcoxon's rank-sum test?- Assesses the difference between two independent samples
- Test whether the populations from which two samples are drawn have the same location
- Again… Functionally the same as the Mann-Whitney test and mimics the design of the independent t-test - 14How do you run a Wilcoxons rank-sum test?LOOK ON NOW LECTUE (MEDIA)
- 15What is the Friedman ANOVA test?- Assesses where the same group of participants complete more than two conditions e.g. Hot v Moderate v cold
- Non parametric version of a one way repeated measures ANOVA - 16Friedmans ANOVA is theory 3, explain it.-Again, Friedman’s ANOVA is based on ranked data
-To begin, the data for your different conditions is placed in different columns
-The data FOR EACH PERSON is then ranked
-E.g. for each person the lowest score will be ranked 1 and so on…
-You then simply add the ranks of each condition - 17How do you run a Friedman ANOVA?LOOK ON NOW LECTURE (MEDIA)
- 18What is the fourth theory?-The Kruskal Wallis test is also based on ranked data
-Therefore, the scores are initially ranked from lowest to highest, ignoring the group they belong to
-Again, you sum the ranks in each group
-The sum of these ranks in each group is denoted by Ri - 19What is the Krusksal Wallis test?-Test of whether more than two independent groups differ
-Non parametric version of one way independent ANOVA - 20How do you run a Kruskal Wallis test?LOOK ON NOW LECTURE (MEDIA)
- 21For the 2-way ANOVA and mixed model ANOVA is there a non parametric test?-For these types of ANOVA there is no non parametric test
-For these parametric tests – data can be normally or not normally distributed
-The test still provides an accurate estimation of the effect
Information
- Description
- Research design & analysis Flashcards on Non-parametric tests, created by Sam Kiernan on 16/04/2019.
- Created by
- Sam Kiernan
- Views
- 5
- Rating