The distributional method for dichotomizing normal data

(Based on Peacock et al 2012, Sauzet et al 2014 and Sauzet et al 2015)

Update 22.01.2016


The distributional method for the dichotomisation of continuous outcome has been developed to allow for the presentation of comparison of proportion after a comparison of mean has been performed (t-test (distdicho) or regression (reg_distdicho)). The method can be applied under the hypothesis that the data is normally distributed with equal variance between the groups (for t-test just two groups can be compared, for regression there is no limit) or with unequal variances. If there are evidence of perturbation to the normal distribution the a skew-normal method can be used (sk_distdicho).


Stata package

A complete package of Stata commands is now available to install directly from Stata. The package has been built with Stata 13 and some command will not work on earlier versions. Please contact me if you need the commands for earlier versions

In stata please type the following:

net describe distdicho, from (

And check regularly for update

adoupdate distdicho


A working paper describing the commands and examples is available here for download



R package

A R-package is also in preparation so check this page for update.



If you have been using the program provided in this page and found bugs or you have any comments concerning them please write to odile dot sauzet at uni bielefeld dot de

Thank you