(I should first say, if "MIDAS" doesn't mean anything to you and
you'd like to know, please see http://gretl.sourceforge.net/midas/
Three cases are relevant when you're using the new --aggr=spread
option to pull high-frequency data into a lower-frequency workspace,
1) The data to be added take the form of a native gretl data file,
with suitable time-series properties.
2) The data are are CSV, but with a suitable time-series structure
that is recognized as such by gretl. (Test: you can use "open" on
the CSV file and gretl gets the time-series information right.)
3) The data are CSV but are not organized such that gretl can "open"
it as time series correctly.
In cases 1) and 2) there shouldn't be any problem: just use
--aggr=spread, as in
# join monthly to quarterly
join hamilton.gdt PC6IT --aggr=spread
# join daily to monthly
join djclose.gdt djclose --aggr=spread
(and similarly for compatible CSV files).
In case 3), however, you'll need to use two options to "join" in
addition to --aggr=spread.
First, use the --tkey option to tell gretl where (and/or how) to
determine the dates of the observations in the file from which
you're adding data (see the chapter on "Joining data sources" in the
Gretl User's Guide for details).
Second, use the (new) --pd option to "join" to tell gretl what the
frequency of the imported data is supposed to be. This option is not
yet in the documentation but that should be fixed before long. A
parameter is required, as in
--pd=12 # monthly
--pd=5 # daily, 5 days per week
(or 6 or 7 for daily data at 6 or 7 days per week). Gretl could
attempt to guess this via (fallible) heuristics, but you as user
should know what you're trying to do; please tell us and then we'll
be sure to get it right.
There follows an example of the syntax in case 3), supposing you
have a monthly data workspace and you wish to import 5-day daily
data from a CSV file in which there's a column headed "obs" holding
dates in the format month/day/4-digit-year:
join mydaily.csv --aggr=spread --tkey="obs,%m/%d/%Y" --pd=5