R language Access Menu

Parametric versus non-parametric tests

Parametric tests are classical statistical test that need the data to have normal distribution. For data that is not normally distributed, there are 2 options. One is to transform the data using functions like log to convert it to normal distribution and then apply parametric test. The second option is to use non-parametric test that do not assume the data to be normally distributed. Following table shows main parametric and non-parametric tests that are done in different situations: 

 

 

              Situation

               Parametric Test

              Non-parametric test

Correlation between 2 numeric  variables

Pearson's correlation

Spearman's or Kendall's correlation

Comparing 2 series of numeric values (paired or unpaired)

Student's t-test

Mann Whitney U test or Wilcoxan test

Comparing multiple groups of numeric values

ANOVA (analysis of variance)

Kruskal-Wallis test

Tables of categorical variables

Chi-squared test

Fisher's Exact test

 

 

 

 


    Comments & Feedback