On Tue, 8 Mar 2016, Sven Schreiber wrote: Perhaps you could give us an idea of how to modify the following test script (from which I cannot provoke an error) so that it's closer to your actual case. <hansl> nulldata 50000 setobs 5 1:1 --stacked-time-series series tul2an = normal() series tucpn = normal() tul2an = uniform() < 0.2 ? NA : tul2an tucpn = uniform() < 0.3 ? NA : tucpn series tully = tul2an - tucpn </hansl> Allin

Am 08.03.2016 um 23:26 schrieb Sven Schreiber: Ok, I think I found the solution -- this is stuff more suitable for the devel list instead of the users list, but in order to finish this thread properly I'll post it here. But first let me note a difference between running gretl in English and in German: When I switched to English gretl gave me a "syntax error" here, where in German there was nothing. So that might be something that needs fixing irrespective of the other thing. The underlying problem was a wrong character. The minus sign in: "series tully = tul2an - tucpn" was actually not a minus sign or hyphen character, but was the en-dash character (I believe U+2013 in Unicode and 0x96 in Windows Western codepage). It came from code by a colleague in a different computer language that I was adapting to hansl. So I think gretl should perhaps check a script if it contains illegal characters outside of comments or literal strings, or when it stumbles over something that it doesn't understand, it could explicitly check if illegal characters have something to do with it and issue a meaningful error message. (Perhaps these characters could also lead to errors like "no such object", when a computer sees a difference between the characters but a human doesn't.) Thanks, sven