With the help of univariate selection, we intend to select single variables that are associated the most with your target variable according to a statistical test.
There are three available tests to base our selection on:
- The f_regression object uses an F-test and a p-value according to the ratio of explained variance against the unexplained one in a linear regression of the variable with the target. This is only useful for regression problems.
- The f_classif object is an ANOVA F test that can be used when dealing with classification problems.
- The Chi2 object is a chi-squared test, which is suitable when the target is a classification and the variables are count or binary data (they should be positive).
All of the tests ...