Non parametrics statistics reading summary

Nonparametric Statistics (Munro , 2005

Nonparametric tests are generally used when the population from which a sample is drawn is not normally distributed . Nonparametric statistics are much easier to calculate manually than their parametric counterparts . Although nonparametric statistics are unable to handle multivariate questions , they are able to analyse data in their original values , thus making interpretation easier

The most commonly used nonparametric test is the chi-square , which is used to compare whether the expected numbers or frequencies of a sample are significantly different from the actual numbers or frequencies . The

chi-square value is calculated by subtracting the expected frequency from the observed frequency for each data value , then squaring the difference and dividing it by the expected frequency , and summing up the result for all the data values . Chi-square is used for nominal (categorical ) data under four main assumptions : frequency data adequate sample size independent measures and theoretical basis for the categorization of the variables

Other nonparametric statistics are the McNemar test , the Man-Whitney U test and the Kruskal-Wallis H test . The McNemar statistic is used to measure changes in two dichotomous measures on the same subjects (i .e dependent variables . The McNemar test uses an adaptation of the chi-square formula to test the direction of the change . Since only the two cells that include changes are included in the analysis , the degrees of freedom are equal to 1 , i .e . 2-1 1

On the other hand , the Man-Whitney U and the Kruskal-Wallis H statistics are used to test independent...