DATA MINING
Desktop Survival Guide by Graham Williams |
|||||
Statistical Tests |
Todo: UNDER CONSTRUCTION
To test whether there is a significant shift in values of a variable for a collection of observations after some event.
Nonparametric tests appropriate when the distribution of variable in is not normally distributed. It might be right skewed or bi-modal, for example.
For small samples (10's) traditional tests for significance that assume a normally distributed sample would not be appropriate.
Wilcoxon Mann-Whitney tests two populations with the null hypothesis being the two populations are similar and the alternative hypothesis being the two populations differ with respect to the median. This is the nonparametric equivalent of the two sample t-test.
Wilcoxon Signed Rank test also compares population medians but uses matched pairs, so it is suitable for analyses of variable before and after an event, for the same entity.