[Gretl-devel] omit --auto and missing values

Allin Cottrell cottrell at wfu.edu
Fri Jul 8 17:30:29 EDT 2016


On Fri, 8 Jul 2016, Sven Schreiber wrote:

> I'm getting "data errors" in a situation where the various series in the 
> dataset have different extents of missing values in them (only at the end or 
> the beginning, so no gappy series [I think]) and I do "omit --auto=0.1".
>
> The errors go away when I precede the estimation and omit commands with a 
> "smpl ... --no-missing" line, that's why I think it has to do with the 
> missings.
>
> Right now I don't have time to cook up a concrete example, but I could try 
> later if this information isn't enough. This is with gretl 2016c on Windows.

An example would be good. I tried cooking up the following, on a 
dataset with n = 51.

<hansl>
open data4-10
INCOME[51] = NA
ols 1 0 2 3 4 5 6 7 8 9
omit --auto=0.1
<hansl>

Here no series have any NAs before I stuck one in for INCOME, which 
I chose because it's the first candidate for omission from the 
kitchen-sink model (P-value 0.7486). So in principle after dropping 
INCOME gretl could estimate the first reduced model using n=51 (as 
opposed to n=50 for the full model), which would then yield an 
invalid comparison and so (hopefully) trigger an error.

However, what actually happened was that gretl kept the original 
n=50 sample as established by the initial OLS: it's not fooled so 
easily. I tried setting INCOME[1] to NA also, for good measure, but 
it's the same story: gretl sticks with the sample defined by the 
original model.

So what you're describing must be something more subtle.

Allin


More information about the Gretl-devel mailing list