Parametric or non-parametic tests?

I know that you use a nonparametric test if the skewness statistic is twice the standard error statistic but in this case do you ignore the sign?For example would a skewness statistic of minus 1.146 (-1.146) be considered twice the size of a standard error of 0.350??? Thanks.


TS Contributor
Generally, if the skewness is large in either direction, a non-parametric test is warranted. That is, consider the absolute value of the skewness.