On Sat, 28 Oct 2006, Rob Tischer wrote: Here is the issue with saving matrices as "ordinary variables" (series): If the matrix has as many rows as there are observations in the data set, or as many rows as in the current sample range, then gretl knows how to align the values from the matrix with the other data series. (In the second case the out-of-sample values are implicitly NA.) Otherwise, the program has no way of knowing how the values should be aligned, and you'll have to help. There are two ways of doing this (without adjusting the sample range), one using a temporary matrix and one using a loop construct. Here's a trivial example of making a series x, of length 20, out of a matrix (column vector) M, of length 10, using a temporary matrix Tmp, with the 10 values in M aligned at the end of the dataset and padded with zeros. (I'm using the syntax for defining matrix expressions that is supported in CVS and the current Windows snapshot): nulldata 20 matrix M = mnormal(10,1) matrix Tmp = zeros(20,1) Tmp[11:20,] = M series x = Tmp delete Tmp print x --byobs Here's the same thing done using a loop: nulldata 20 matrix M = mnormal(10,1) series x = 0 loop i=11..20 x[$i] = M[$i-10,1] end loop print x --byobs You could actually do the same thing using the smpl command too: nulldata 20 matrix M = mnormal(10,1) series x = 0 smpl 11 20 x = M smpl --full print x --byobs Allin Cottrell