Re: [Gretl-users] Question about join
On Mon, 8 Dec 2014, Allin Cottrell wrote:
On Tue, 9 Dec 2014, Logan Kelly wrote:
> Thanks Allin, here is an example:
>
> string frequency = "q"
> #join "gretl_tmp_3785.csv" out
--tkey="obs,%Y@frequency%@frequency"
> --aggr="avg" --data="ENU2705340010"
> join "gretl_tmp_3785.csv" out --tkey="obs,%Yq%q"
--aggr="avg"
> --data="ENU2705340010"
>
> Both join commands yields the same error. Attached is the csv file (also
> pasted below).
>
> obs, ENU2705340010
> 2013q4, 1209
> 2013q3, 1163
> 2013q2, 1142
> 2013q1, 1276
> 2012q4, 1239
> 2012q3, 1135
> 2012q2, 1123
> 2012q1, 1278
Thanks, Logan. Now what does the "inner" dataset look like before you try the
join? For comparison, here's an example I just tried:
I open data9-7.gdt (quarterly data 1975:1 to 1990:4) and then I try using
join on the following data file, named "import.csv":
obs x
1980q1 1
1980q2 2
using this script:
<hansl>
open data9-7.gdt
join import.csv x --tkey="obs,%Yq%q"
print QNC x -o
</hansl>
This works as expected, showing values 1 and 2 for "x" in quarters
1980:1 and 1980:2 respectively.
OK, I think I can guess what's happening. You're using the option
--aggr=avg so I suppose you're trying to average quarterly data into
an annual dataset. That won't work at present via join unless you
have an annual key variable. I'll have to think about that (and I
admit, the error message you saw is misleading). But in the meantime
the following works.
My source file, import2.csv:
year quarter x
1980 1 1
1980 2 2
1980 3 3
1980 4 4
My script:
<hansl>
open data3-6.gdt # annual data
join import2.csv x --tkey="year" --aggr=avg
print Ct x -o
</hansl>
Allin