On Tue, 19 Mar 2019, Boris Demeshev wrote:

> Have compiled gretl the current snapshot. It indeed loads the data
> well!!!! :) :)
>
> But!
>
> When I try to assign panel data structure (ID_W for weave-time and
> ID_I or REDID_I for individual) the app is killed once again :(
> Message in terminal:
> pad_panel_dataset: n*T = 65423*22 = 1439306 but dset->n = 335594
> Killed

Out of memory. In gretl a panel dataset must be "nominally balanced",
or in other words have an equal number of observation slots for each
individual, even if some of those slots are just padded with NAs.

This dataset is pretty sparse (it's longitudinal but not really a
panel as such), and padding it out to a nominally balanced panel is
going to increase its size by a factor of 1439306 / 335594 ~= 4.3. If
the data occupy about 7 GB to start with this will not be feasible
unless you have a ton of RAM (well over 32 GB I guess).

I think you'll have to do some clever subsetting of the data to enable
estimation.

Allin
_______________________________________________
Gretl-devel mailing list
Gretl-devel@lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-devel