I'm doing a multiple regression analysis and my outcome variable is strongly positively skewed.

Without transformation skewness = 2.10 and kurtosis = 5.11 and the Q-Q plot looks bad

with a log-transformation skewness = 1.44 and kurtosis = 1.67 and the Q-Q plot looks better but still not very good

An inverse transformation gives skewness=.99 and kurtosis=-.01 and now the Q-Q plot looks much better as do the residuals after the regression. Thus I want to use the inverse transformation.

My question:

the original range of the variable is 5-14 and with the inverse transformation it becomes

.07 - .20

However the sign of the regression coefficient changes so I want to reverse the distribution, how is this correctly done? Can I do it after the inverse transformaion by multiplying with -1 and adding 1.20 ?

Thanks, Annette