On Fri, 24 Apr 2015, Henrique Andrade wrote: The answer is No. The rule with gretl functions is that they cannot have arbitrary side effects. They can return a value, and they can change the values of arguments given in "pointer" form, but that's it. They can't mash up your dataset. Notice that in your example, the series argument is bogus: it implies that something is being done with a particular series, but if the "dataset" command were allowed in a function it would actually transform the entire dataset. It's possible to do something that's related, without breaking the rules. If a function is given a series or list argument, you can turn that into a matrix, then process the matrix however you want, and return the result as a matrix. Thus, given a quarterly series one could return an annualized version in the form of a column vector. Allin