![[logo] US EPA](http://www.epa.gov/epafiles/images/logo_epaseal.gif){kind=logo}
Comparisons of Statistical Tests
Links:
Statistical methods can be grouped roughly into 3 categories: descriptive, inferential and exploratory. Descriptive statistics summarize observations in terms of central tendency, i.e. the mean, or dispersion, i.e. the variance. When we perform a statistical test and draw conclusions based on the results we are using inferential statistics. Some methods are exploratory, such as canonical correlation, and don't have p-values associated with them, but are designed to look for patterns in the data.
Nonparametric Equivalent Tests
This table provides a list of nonparametric equivalent tests for common univariate and bivariate tests. Nonparametric tests are appropriate when the data are not necessarily distributed normally, that is, the data may not be unimodal or symmetric. A common application for nonparametric testing occurs when a response variable has a few very large values which skew the distribution to the right.
Comparison of Methods
A good way to understand how a particular statistical method works is by comparing it to a similar or related method. Each of the statistical methods presented in this site is paired with every other method and the similarities and differences are described for each pair.
Multivariate Model Comparisons
The differences between multivariate methods are presented in terms of the underlying number and types of independent (or “predictor”) variables and the number and types of dependent (or “response”) variables.