Dear Forum
I have normal as well as skewed data comparing 2 groups and I would like to report standardized mean differences (SMD). I would very much appreaciate advice on how to handle the nonparametric data.
Meantime, is it reasonable to log transform (var >>>> ln_var) and report SMD on the resultant data? If so, do the SMDs - as a form of standardized effect size - retain the same interpretation when comparing the 2 groups? that is, scale for SMD: <0.10 indicates negligible; <0.20 indicates small; 0.20 to 0.39 indicates small-to-moderate; 0.40 to 0.59 indicates moderate; 0.60 to 0.79 indicates moderate-to-large; and >0.80 indicates large?
Secondly, if the above is a reasonable approach, what summary estimate for the variable (var) would I report (eg, in Table 1): mean of log transformed variable (ln_var) ? Or exponentiated mean of ln_var?
Thanks in advance.
Itai
I have normal as well as skewed data comparing 2 groups and I would like to report standardized mean differences (SMD). I would very much appreaciate advice on how to handle the nonparametric data.
Meantime, is it reasonable to log transform (var >>>> ln_var) and report SMD on the resultant data? If so, do the SMDs - as a form of standardized effect size - retain the same interpretation when comparing the 2 groups? that is, scale for SMD: <0.10 indicates negligible; <0.20 indicates small; 0.20 to 0.39 indicates small-to-moderate; 0.40 to 0.59 indicates moderate; 0.60 to 0.79 indicates moderate-to-large; and >0.80 indicates large?
Secondly, if the above is a reasonable approach, what summary estimate for the variable (var) would I report (eg, in Table 1): mean of log transformed variable (ln_var) ? Or exponentiated mean of ln_var?
Thanks in advance.
Itai
