Re: [Gretl-devel] Time series data with gaps and plots
On Wed, 25 Oct 2017, Sangvinatsos, Antonios wrote:
The test data you provided below worked fine. Gretl did not ask me
to give it a time series of panel interpretation and immediately
understood it is a time series.
OK, that's good.
I suspected that I perhaps have to change the name of my date
variable from ASOF_DATE to DATE.
There's a short list of column labels that gretl takes to mean that
the first column is an observation label or date, and not regular
data. This includes "obs", "observation", "date" (in lower
or upper
case) or an empty cell. See Chapter 4, "Data Files", in the Gretl
User's Guide.
When I do this the program understands the first variable as a row
label variable (improvement) but it does not understand it as a date
[...]
gretl says:
line: DATE,MKT
scanning for row labels and data...
first row label "10/1/2007", last label "9/25/2017"
There's the problem: the format here is not consistently MM/DD/YYYY:
days and months should be zero-padded, as in "09" or "01". One test
that gretl applies for consistent date strings is that they should all
be of the same length, and that won't be the case without
zero-padding.
However, the requirement of exact consistency is now relaxed in gretl
git; we'll try to handle a bit of sloppiness, allowing the length of
the date strings to differ by one character. We could try relaxing
this a little further.
Allin