Non Parametric Tests Inferential Analysis Flashcards
How to choose the correct statistical test
1) Is it a test of difference, a correlation or an association?
2) What design have I used? Independent measures or repeated measures/matched participants design?
3) What level of measurement have I used? Nominal, ordinal, interval, ratio?
4) Is the test non parametric or parametric?
Are my results significant or not? Do they show a real difference or not?
You need to compare results gained from a statistical test to critical values (that can be found in the appendix of your textbook). Please read the information about each of the statistical tests carefully because it summarizes whether your results need to be greater or smaller than the critical values in the table in order to be significant or not.
What are critical values
When working out the statistics for your research (by using one of the four statistical tests) you will compare your final result with a critical value. A critical value is a numerical value found in the statistical tables in the appendix that helps to determine the significance of your results. It can determine whether your results are significant or not – for instance is there a real difference between two conditions (if yes, then the results are significant and the experimental hypothesis should be accepted), or is the difference due to fluke or chance (if this is true, then we accept the null hypothesis).
Spearman’s rho
• This is used when we wish to test a relationship or correlation between two variables.
• The data should be ordinal (at least).
• The data are in related pairs
• This is a non-parametric test
• Spearman’s Rho calculates the strength of the relationship between the two variables e.g. if r is -1.0 this means there is a perfect negative correlation between two variables, and if r is +1.0 then this is a perfect positive correlation.
Mann Whitney
• This test is used when we are looking for a test of difference
• This statistical test is used when we have an independent group design.
• We also use this test when we have data that is ordinal (can be placed in rank order) or interval.
• This test is a non-parametric test
Chi squared
• This test can be used for either a test of association or for a test of difference (that is why this is a special statistical test).
• We use chi squared (for a test of difference) when we have an independent measures design and are using nominal data in the form of frequencies or separate categories.
• We would also use chi squared if we are testing an association that might occur between two variables. Again, the data would be nominal and would use an independent measures design
• Chi squared is a non-parametric test
Wilcoxon
• It is used to measure a test of difference
• This test is used when a repeated measures design or matched participants design has been used.
• The data must be ordinal
• Wilcoxon is a non-parametric test
When using inferential (statical) tests, remember the following points
1) Justify your choice of statistical test (by referring to whether it is a test of difference, a correlation or an association, what design has been used, and also the level of measurement used.) Refer to page 738
2) Calculate the results of the statistical test.
3) Compare the results to the critical value (are they significant or not?)
4) Please mention the level of significance that has been selected e.g. p<0.05 (5%) and whether the hypothesis was one tailed or two tailed.
5) The end result is very important. Have you accepted the experimental hypothesis or rejected it? What about the null hypothesis? Has this been accepted or rejected? (Please see examples on page 739 about how to accurately report the results)
