[Gretl-devel] Questions about the "dataset resample n" command

[Allin Cottrell]

On Tue, 4 Mar 2008, andreas.rosenblad at ltv.se wrote:

> cottrell at wfu.edu @ INTERNET skrev 2008-03-04 03:35:17 :
> 
> > On Mon, 3 Mar 2008, andreas.rosenblad at ltv.se wrote:
> >
> > > After using the command "dataset resample n" to resample >
> > observations with replacement, how does one restore the original >
> > full data set?
> >
> > You can't.  This is not documented yet because I'm not yet sure that
> > it's useful, or should be retained.
> 
> OK. I think the function is very useful and should be retained. 
> It make bootstrapping of all statistical methods implemented in 
> gretl very easy. But to be fully usable there must of course be 
> a command to restore the full data set. I hope it will be 
> implemented as well.

The difficulty is that "restoration of the full data set" could 
not mean the same in this case as in the other cases of sampling.

In the other cases, we can "carry back" both modifications to 
existing series that were made when sub-sampled and series that 
were newly created.  There's no way to do this with a resampled 
dataset because there's no unique mapping from rows in the 
modified dataset to rows in the original.  We could preserve 
scalars and matrices newly modified or created, but we'd have to 
discard any changes to series.

With that proviso, it could be done.  

Allin.