hi
i have a large data set which is negative skewed- i can
transform the data using ln (constant - dependent variable)
and will then be testing for differences in medians
i would like to do this using anova / factorial anova where
appropriate
my problem is that when i run anova (simple) on the
transformed variable - i get levenes statistic for unequal
variances which is still significant, i assume i use a test
of differences in medians such as tahene's
i would also like to back transform to the grouped median
(there are a considerable number of ties in the original
variable) but i do not get the same answer as in the
untransformed data
also i am using SPSS - does anyone know how the grouped
median is calculated, i can get close
regards
ann
----------------------
ann e fitzmaurice
medical statistician
dugald baird centre
university of aberdeen
[log in to unmask]
tel 01224 681818 extension 53876
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|