On Mon, Nov 20, 2023 at 7:56 PM Cottrell, Allin <cottrell(a)wfu.edu&gt; wrote: And also a non-trivial piece of work, I see. With the original data taking the form of an old-school fixed-format file plus codebook, and an erroneous "unique ID" to be spotted and fixed. Kudos. Allin

I commented on Jack's Card and Krueger dataset for gretl: Follow-up: The original C&K numerical data are in a plain text file called public.dat, with no variable names and with "." indicating NA. It struck me that gretl ought to be able to read such a file, but it couldn't, on two counts (see below). I've now fixed things in git so that it works. Problem #1:Given the ".dat" suffix, gretl was expecting a JMulTi data file, and on that interpretation the read failed right away. But ".dat" is pretty generic, so now we don't jump to conclusions; we run a quick test of the file's content and if it doesn't validate as JMulTi we treat it as generic plain text ("CSV" in an extended sense). Problem #2: Gretl scans for variable names on the first line of a "CSV" file. It's not going to find any in public.dat, and in itself that's OK, but the scanner was being freaked out by the "." cells, which seemed to be neither numeric nor interpretable as variable names. The update here is that on the varname scan we treat plain "." as indicating NA, which can be taken as numeric. So now the scan completes successfully (not finding any names), and gretl adds generic variable names "v1", "v2" and so on. (On the actual data read, as opposed to the varname scan, we were already treating "." as NA.) Allin