Hello,
I have two studies with involving percepntage data. One has a significant
skew value, and one has a significant kurtosis value (I calculated these by
dividing the values by their respective standard errors, and then checking
whether the value was greater than the critial value of 1.96, as mentioned by
Field (2005)). Field suggests that unevenly distributed data can be
transformed, although for negative skew (which my data has), you should
reverse the data first, by subtracting each value from the highest observed
value. I did this, and then square-rooted the data, and it appears to have
improved the distribution. However, since my transformed data is reversed,
large values are now small and small values are now large, which makes things
difficult to interpret. Is there any way to 're-reverse' the transformed variable,
or do I just interpret it as it is (backwards, essentially).
Thanks a lot for your help.
|