Re: Multiple linear regression: How much deviation from normal is too much?


"However, I lose the wide
range of scores in the original data, which goes from 0 to 18 to 0 to
4.24. Because the original instrument is based on a 0-24 scale, a range

of 0 to 4.24 distorts the true prevalence of corporal punishment among
my sample as measured by the instrument."

No. The scale has changed. If you divide by 60 to go from minutes to
hours you don't lose range in any meaningful sense. This is not that
dissimilar. You can reverse the transformation for
reporting/interpreting purposes. For a log transformation this gives
the geometric mean rather than the arithmetic mean.

Thom

.