Response Transformation: An Example

Authors


ABSTRACT

A set of data involving 108 observations from a previous study on a sterilized, concentrated baby formula is reconsidered. Originally, four separate second degree models involving 84 different coefficients were fitted. However, if the response values are transformed from y to In y, a model containing just five parameters explains 87% of the variation in all of the data about their grand mean. Moreover, this simple equation does not include one of the five predictor variables, a fact not previously apparent. These data provide an example of the effectiveness of a transformation.

Ancillary