Transformations (ladder of powers)

Transformation deals with non-normality of the data points and non-homogeneous variance. The power transformations form the following ladder:..., ..... Provided x > 1, powers below 1 reduce the high values relative to the low values as in positively skewed data, whereas, powers above 1 have the opposite effect of stretching out high values relative to low ones, as in negatively skewed data. All power transformations are monotonic when applied to positive data (they are either increasing or decreasing, but not first increasing and then decreasing, or vice versa). The square root transformation often renders Poisson data approximately normal.