On Sun, May 16, 2021 at 11:00:44AM -0700, Ben Pfaff wrote:
Yes, I agree. Classic SPSS isn't general purpose enough to write
statistical procedures
that are as easy to use as the ones built into it. The SPSS language
manages to
be a misery of inconsistencies that make it near impossible to generalize.
The macro language (which I'm currently implementing), which appears to be
meant for extensions, is terrible.
Maybe we will eventually be able to implement the Python extensions to
SPSS.
Those are the most fruitful direction I've seen toward making SPSS
programmable
in a reasonably friendly way.
Some years ago I wrote an experimental scheme interface which seemed to work
quite well.
Perhaps I'll dig it up again some time. The biggest complication as I remember
was
dealing with missing values. They always complicate matters in unexpected ways.
J'
