On 30 April 2015 at 17:51, Michael Ott <ottmichael(a)gmail.com&gt; wrote: Does anyone know stepwise how to do an interrupted time series in Gretl. If you've yet to receive an answer about this, it's because most practioners believe stepwise regression of any sort is a Very Bad Idea. Google Frank Harrell's 11 points against the method to understand why, If you've read this already, then you'll need to give us a darned good explanation as to why you think he's wrong. If your choice of predictors are guided by theory, what exactly is wrong with entering the model as a whole anyway?