Parcel scale or aggregation

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Parcel scale or aggregation

John Morgan
Hello, I am working on preparing data for to run in a spatial
autoregression model (probably SEM). You are one of the most well
grounded people I am aware of in this type of methodology. I am hoping you can
help me with a simple question.

My question is this: we have the data at the scale of the parcel (or
household) and there are a couple of hounded thousand records across the
size of a large metropolitan area. When feeding the data into the
model, is there
a reason/requirement to aggregate our data variable to some boundary scale
such as a city block? Or is it ok to keep it at the parcel scale? We are
interested in analyzing characteristics at e.g. household level similar to
a hedonic model.

Thanks for any feedback.

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Parcel scale or aggregation

Dexter Locke
Hello.

This is more of a research design question than a spatial analysis question.

If you research question pertains to parcels and you have parcel data, then
why aggregate?

Neighborhood-level attributes can be important. They can be included by
attributing parcels with their neighborhood characteristics, with random
effects at the neighborhood scale, among other techniques.

The chosen method depends on the research questions.

-Dexter



On Tue, Apr 21, 2020 at 9:34 AM John Morgan <[hidden email]> wrote:

> Hello, I am working on preparing data for to run in a spatial
> autoregression model (probably SEM). You are one of the most well
> grounded people I am aware of in this type of methodology. I am hoping you
> can
> help me with a simple question.
>
> My question is this: we have the data at the scale of the parcel (or
> household) and there are a couple of hounded thousand records across the
> size of a large metropolitan area. When feeding the data into the
> model, is there
> a reason/requirement to aggregate our data variable to some boundary scale
> such as a city block? Or is it ok to keep it at the parcel scale? We are
> interested in analyzing characteristics at e.g. household level similar to
> a hedonic model.
>
> Thanks for any feedback.
>
>         [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-Geo mailing list
> [hidden email]
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Parcel scale or aggregation

John Morgan
Dexter.

Your point that it is more of a research design question is well taken,
but I figured if anyone would know it would be a group of folks using R
for spatial analysis. The research question does pertain to the parcels
and the explanatory variables will be computed at that scale. I just
didn't know if that matter per say for functions like sarlm() or or
gmerrorsar(). It sounds like not so much as they will be treat the same
as rows in a data frame regardless of what scale they are at.

Thanks,

Derek

On 4/21/2020 9:59 AM, Dexter Locke wrote:

> Hello.
>
> This is more of a research design question than a spatial
> analysis question.
>
> If you research question pertains to parcels and you have parcel data,
> then why aggregate?
>
> Neighborhood-level attributes can be important. They can be
> included by attributing parcels with their neighborhood
> characteristics, with random effects at the neighborhood scale, among
> other techniques.
>
> The chosen method depends on the research questions.
>
> -Dexter
>
>
>
> On Tue, Apr 21, 2020 at 9:34 AM John Morgan <[hidden email]
> <mailto:[hidden email]>> wrote:
>
>     Hello, I am working on preparing data for to run in a spatial
>     autoregression model (probably SEM). You are one of the most well
>     grounded people I am aware of in this type of methodology. I am
>     hoping you can
>     help me with a simple question.
>
>     My question is this: we have the data at the scale of the parcel (or
>     household) and there are a couple of hounded thousand records
>     across the
>     size of a large metropolitan area. When feeding the data into the
>     model, is there
>     a reason/requirement to aggregate our data variable to some
>     boundary scale
>     such as a city block? Or is it ok to keep it at the parcel scale?
>     We are
>     interested in analyzing characteristics at e.g. household level
>     similar to
>     a hedonic model.
>
>     Thanks for any feedback.
>
>             [[alternative HTML version deleted]]
>
>     _______________________________________________
>     R-sig-Geo mailing list
>     [hidden email] <mailto:[hidden email]>
>     https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>     <https://stat.ethz.ch/mailman/listinfo/r-sig-geo>
>
--
John Derek Morgan, Ph.D., GISP
Assistant Professor of GIS
Earth & Environmental Sciences
University of West Florida
https://uwf.edu/go/gis/
https://pages.uwf.edu/jmorgan3
>>>Chat directly with me via Google Chat our Hangout<<<


        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Parcel scale or aggregation

Dexter Locke
Yes, those functions should work provided your data are shaped correctly
(they sound like they are).

I imagine the results will be sensitive the chosen spatial weights matrix.
That choice indirectly gets to your point about aggregation. Some types of
weights may include spatial neighbors that might assimilate neighborhoods
or blocks.

Good luck!

-Dexter
http://dexterlocke.com/


On Tue, Apr 21, 2020 at 11:23 AM derek <[hidden email]> wrote:

> Dexter.
>
> Your point that it is more of a research design question is well taken,
> but I figured if anyone would know it would be a group of folks using R for
> spatial analysis. The research question does pertain to the parcels and the
> explanatory variables will be computed at that scale. I just didn't know if
> that matter per say for functions like sarlm() or or gmerrorsar(). It
> sounds like not so much as they will be treat the same as rows in a data
> frame regardless of what scale they are at.
>
> Thanks,
>
> Derek
> On 4/21/2020 9:59 AM, Dexter Locke wrote:
>
> Hello.
>
> This is more of a research design question than a spatial
> analysis question.
>
> If you research question pertains to parcels and you have parcel data,
> then why aggregate?
>
> Neighborhood-level attributes can be important. They can be included by
> attributing parcels with their neighborhood characteristics, with random
> effects at the neighborhood scale, among other techniques.
>
> The chosen method depends on the research questions.
>
> -Dexter
>
>
>
> On Tue, Apr 21, 2020 at 9:34 AM John Morgan <[hidden email]> wrote:
>
>> Hello, I am working on preparing data for to run in a spatial
>> autoregression model (probably SEM). You are one of the most well
>> grounded people I am aware of in this type of methodology. I am hoping
>> you can
>> help me with a simple question.
>>
>> My question is this: we have the data at the scale of the parcel (or
>> household) and there are a couple of hounded thousand records across the
>> size of a large metropolitan area. When feeding the data into the
>> model, is there
>> a reason/requirement to aggregate our data variable to some boundary scale
>> such as a city block? Or is it ok to keep it at the parcel scale? We are
>> interested in analyzing characteristics at e.g. household level similar to
>> a hedonic model.
>>
>> Thanks for any feedback.
>>
>>         [[alternative HTML version deleted]]
>>
>> _______________________________________________
>> R-sig-Geo mailing list
>> [hidden email]
>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>
> --
> John Derek Morgan, Ph.D., GISP
> Assistant Professor of GIS
> Earth & Environmental Sciences
> University of West Floridahttps://uwf.edu/go/gis/https://pages.uwf.edu/jmorgan3
> >>>Chat directly with me via Google Chat our Hangout<<<
>
>

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo