For variables that were not normally distributed, scores were transformed using the square root and natural
logarithmic transformations.
Square root and
logarithmic transformations were used for all variables that were positively skewed.
In all cases, square root transformations were attempted first, and when such transformations failed to create normal distributions,
logarithmic transformations were used.
We thus transformed each of these variables prior to analysis (using square root and
logarithmic transformations as each case required).
When variables were negatively skewed, the scale was first reflected and then transformed using a square root or
logarithmic transformation.
Not exact matches
Because both parents» internalizing problems were positively skewed,
logarithmic (log10)
transformations of scores were used to approach normal distributions (Tabachnick and Fidell 2012).