On Sat, 19 Dec 2020, Sven Schreiber wrote: Same thing outside of a function. If you're subsampled and want to store a matrix as a series inside a bundle you have to make it into a proper series first. That's the only way gretl can know where it's supposed to start; there's no notion of a dataset inside a bundle. However, if the matrix (column vector) in question is produced by a gretl function that sets t1 and t2 on it, best practice would be to store it in the bundle as a matrix, then pull it out as a series when it's needed. Here's a sketch of how one might get matrix -> series alignnment right when the matrix in question doesn't have t1 and t2 set automatically. <hansl> function series extract_series (const bundle bx) smpl bx.t1 bx.t2 return bx.mat end function open data3-6 bundle b # will contain more stuff? smpl 1965 1990 matrix x = mnormal($nobs, 1) bundle b.bx = _(mat=x, t1=$t1, t2=$t2) smpl full # do stuff... series s = extract_series(b.bx) print s -o </hansl> Allin