[R-sig-Geo] lm.morantest

[Dexter Locke]

Hi Ana,

Maybe you need to use the zero.policy = T option when creating your weights
matrix? Alternatively, consider a different style of weights that will
surely create connections.

Since your data are multi-level, consider also the ICC (Intraclass
Correlation Coefficient).

The Moran's I test of the residuals and ICC will help determine the degree
of spatial autocorrelation, and hierarchical correlation, respectively.

HTH,
Dexter



On Mon, Aug 22, 2016 at 11:13 AM, Anita Morales <anita.morales at gmail.com>
wrote:

> Dear all,
>
> I'm relative new to the list, being this my first question. I have a
> problem when trying to estimate a Moran test for residuals in the second
> level when there are missing areas. My dataset contains information on
> individual nested in cities, the shapefile contains information of the 342
> cities, while my dataset contains individuals that belong to 288 cities.
>
> I've used multilevel models to estimate area level variation and now I want
> to test the autocorrelation for the residuals in the second level (cities).
> I'm using "lm.morantest"; however, I'm getting an error about "areas with
> no neighbours" (I already checked that and I'm quite sure that all my areas
> are connected). I think the problem may be caused by the missing data
> (residuals) for some areas. Since the listw object created, based on the
> shape file, contains neighbourhood weights for 342 cities, while from the
> estimated multilevel model, I've got residuals data for only 288 cities.
>
> Could you please advise me as to how to test for autocorrelation in the
> residuals when there are missing data?
> --
> Ana
>
>         [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>

	[[alternative HTML version deleted]]