mantel correlogram

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

mantel correlogram

Tyler Smith-2
Hello,

We're preparing a field sampling program, and would like to determine
a minimum distance between samples to reduce/eliminate spatial
autocorrelation. I think a good approach would be to calculate a
mantel correlogram, and use the range of the correlogram as our
minimum sampling distance.

* Questions

1) is this a reasonable approach
2) if so, how best to do this?

* Details
We have a vector map with the point coordinates of several hundred
potential sampling sites, and ~ 10 raster layers with appropriate data
to test for spatial autocorrelation (WORLDCLIM, soils). I could do
something like the following, but I'm not sure if there's a simpler or
more appropriate approach:

1) extract the raster data for each point
2) save the data to csv; import into R
3) calculate the spatial distances between points, after projecting
the lat-long data into an appropriate scale (?)
4) calculate the climate distance using the WORLDCLIM data
5) use the 'mgram' function in the 'ecodist' package to calculate the
actual correlogram between the spatial distance and climate distance

Any suggestions on the approach or the methods would be welcome!

Thanks,

Tyler
_______________________________________________
grass-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/grass-user
Reply | Threaded
Open this post in threaded view
|

Re: mantel correlogram

Alex Mandel-2
Use R. It includes Moran's I and Geary's C tests for
spatial-autocorrelation. Look like it has  mantel too.

You'll probably need the sp, spdep and rgdal packages. You might also
want to use the Raster package to extract the sampling data, or you can
use spGRASS to tie the R and Grass together.

See chapter 9 (1st ed) of Applied Spatial Data Analysis with R.
http://www.asdar-book.org/

Enjoy,
Alex

On 03/11/2014 09:18 AM, Tyler Smith wrote:

> Hello,
>
> We're preparing a field sampling program, and would like to determine
> a minimum distance between samples to reduce/eliminate spatial
> autocorrelation. I think a good approach would be to calculate a
> mantel correlogram, and use the range of the correlogram as our
> minimum sampling distance.
>
> * Questions
>
> 1) is this a reasonable approach
> 2) if so, how best to do this?
>
> * Details
> We have a vector map with the point coordinates of several hundred
> potential sampling sites, and ~ 10 raster layers with appropriate data
> to test for spatial autocorrelation (WORLDCLIM, soils). I could do
> something like the following, but I'm not sure if there's a simpler or
> more appropriate approach:
>
> 1) extract the raster data for each point
> 2) save the data to csv; import into R
> 3) calculate the spatial distances between points, after projecting
> the lat-long data into an appropriate scale (?)
> 4) calculate the climate distance using the WORLDCLIM data
> 5) use the 'mgram' function in the 'ecodist' package to calculate the
> actual correlogram between the spatial distance and climate distance
>
> Any suggestions on the approach or the methods would be welcome!
>
> Thanks,
>
> Tyler
> _______________________________________________
> grass-user mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/grass-user
>

_______________________________________________
grass-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/grass-user
Reply | Threaded
Open this post in threaded view
|

Re: mantel correlogram

Thomas Adams-2
Alex,

I believe Tyler does plan on using R for the statistical analyses, but using GRASS GIS in combination with R is the easiest path, I think.

Tom

On Tuesday, March 11, 2014, Alex Mandel <[hidden email]> wrote:
Use R. It includes Moran's I and Geary's C tests for
spatial-autocorrelation. Look like it has  mantel too.

You'll probably need the sp, spdep and rgdal packages. You might also
want to use the Raster package to extract the sampling data, or you can
use spGRASS to tie the R and Grass together.

See chapter 9 (1st ed) of Applied Spatial Data Analysis with R.
http://www.asdar-book.org/

Enjoy,
Alex

On 03/11/2014 09:18 AM, Tyler Smith wrote:
> Hello,
>
> We're preparing a field sampling program, and would like to determine
> a minimum distance between samples to reduce/eliminate spatial
> autocorrelation. I think a good approach would be to calculate a
> mantel correlogram, and use the range of the correlogram as our
> minimum sampling distance.
>
> * Questions
>
> 1) is this a reasonable approach
> 2) if so, how best to do this?
>
> * Details
> We have a vector map with the point coordinates of several hundred
> potential sampling sites, and ~ 10 raster layers with appropriate data
> to test for spatial autocorrelation (WORLDCLIM, soils). I could do
> something like the following, but I'm not sure if there's a simpler or
> more appropriate approach:
>
> 1) extract the raster data for each point
> 2) save the data to csv; import into R
> 3) calculate the spatial distances between points, after projecting
> the lat-long data into an appropriate scale (?)
> 4) calculate the climate distance using the WORLDCLIM data
> 5) use the 'mgram' function in the 'ecodist' package to calculate the
> actual correlogram between the spatial distance and climate distance
>
> Any suggestions on the approach or the methods would be welcome!
>
> Thanks,
>
> Tyler
> _______________________________________________
> grass-user mailing list
> <a href="javascript:;" onclick="_e(event, &#39;cvml&#39;, &#39;grass-user@lists.osgeo.org&#39;)">grass-user@...
> http://lists.osgeo.org/mailman/listinfo/grass-user
>

_______________________________________________
grass-user mailing list
<a href="javascript:;" onclick="_e(event, &#39;cvml&#39;, &#39;grass-user@lists.osgeo.org&#39;)">grass-user@...
http://lists.osgeo.org/mailman/listinfo/grass-user


--
Thomas E Adams, III
718 McBurney Drive
Lebanon, OH 45036

1 (513) 739-9512 (cell)



_______________________________________________
grass-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/grass-user
Reply | Threaded
Open this post in threaded view
|

Re: mantel correlogram

Alex Mandel-2
You are right, I didn't read it that closely 1st time around. My point
was that all of it can be done in R, and there are geospatial specific
packages that have all the tests one might want. The bare minimum
interaction is via rgdal or spgrass to pull data over from existing
GRASS data sets. If the data isn't already in GRASS then rgdal one can
read the original files directly. No need to pass csv around. Of course
if it is in GRASS then you should have it all in the same projection
already anyways if you put it all into the same mapset/location.

The other book likely to have exactly what you want (field sampling
design) is Ch 5.
http://www.amazon.com/Spatial-Analysis-Ecology-Agriculture-Using/dp/1439819130/ref=la_B001K6MGR8_1_1?s=books&ie=UTF8&qid=1394557436&sr=1-1

Enjoy,
Alex

On 03/11/2014 09:58 AM, Thomas Adams wrote:

> Alex,
>
> I believe Tyler does plan on using R for the statistical analyses, but
> using GRASS GIS in combination with R is the easiest path, I think.
>
> Tom
>
> On Tuesday, March 11, 2014, Alex Mandel <[hidden email]> wrote:
>
>> Use R. It includes Moran's I and Geary's C tests for
>> spatial-autocorrelation. Look like it has  mantel too.
>>
>> You'll probably need the sp, spdep and rgdal packages. You might also
>> want to use the Raster package to extract the sampling data, or you can
>> use spGRASS to tie the R and Grass together.
>>
>> See chapter 9 (1st ed) of Applied Spatial Data Analysis with R.
>> http://www.asdar-book.org/
>>
>> Enjoy,
>> Alex
>>
>> On 03/11/2014 09:18 AM, Tyler Smith wrote:
>>> Hello,
>>>
>>> We're preparing a field sampling program, and would like to determine
>>> a minimum distance between samples to reduce/eliminate spatial
>>> autocorrelation. I think a good approach would be to calculate a
>>> mantel correlogram, and use the range of the correlogram as our
>>> minimum sampling distance.
>>>
>>> * Questions
>>>
>>> 1) is this a reasonable approach
>>> 2) if so, how best to do this?
>>>
>>> * Details
>>> We have a vector map with the point coordinates of several hundred
>>> potential sampling sites, and ~ 10 raster layers with appropriate data
>>> to test for spatial autocorrelation (WORLDCLIM, soils). I could do
>>> something like the following, but I'm not sure if there's a simpler or
>>> more appropriate approach:
>>>
>>> 1) extract the raster data for each point
>>> 2) save the data to csv; import into R
>>> 3) calculate the spatial distances between points, after projecting
>>> the lat-long data into an appropriate scale (?)
>>> 4) calculate the climate distance using the WORLDCLIM data
>>> 5) use the 'mgram' function in the 'ecodist' package to calculate the
>>> actual correlogram between the spatial distance and climate distance
>>>
>>> Any suggestions on the approach or the methods would be welcome!
>>>
>>> Thanks,
>>>
>>> Tyler

>

_______________________________________________
grass-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/grass-user
Reply | Threaded
Open this post in threaded view
|

Re: mantel correlogram

Tyler Smith-2
Thanks for your suggestions. It looks like the R Borg is continuing to assimilate procedures that once required specialty software. Time to learn some new packages!

Tyler

On March 11, 2014 1:08:10 PM EDT, Alex Mandel <[hidden email]> wrote:

>You are right, I didn't read it that closely 1st time around. My point
>was that all of it can be done in R, and there are geospatial specific
>packages that have all the tests one might want. The bare minimum
>interaction is via rgdal or spgrass to pull data over from existing
>GRASS data sets. If the data isn't already in GRASS then rgdal one can
>read the original files directly. No need to pass csv around. Of course
>if it is in GRASS then you should have it all in the same projection
>already anyways if you put it all into the same mapset/location.
>
>The other book likely to have exactly what you want (field sampling
>design) is Ch 5.
>http://www.amazon.com/Spatial-Analysis-Ecology-Agriculture-Using/dp/1439819130/ref=la_B001K6MGR8_1_1?s=books&ie=UTF8&qid=1394557436&sr=1-1
>
>Enjoy,
>Alex
>
>On 03/11/2014 09:58 AM, Thomas Adams wrote:
>> Alex,
>>
>> I believe Tyler does plan on using R for the statistical analyses,
>but
>> using GRASS GIS in combination with R is the easiest path, I think.
>>
>> Tom
>>
>> On Tuesday, March 11, 2014, Alex Mandel <[hidden email]>
>wrote:
>>
>>> Use R. It includes Moran's I and Geary's C tests for
>>> spatial-autocorrelation. Look like it has  mantel too.
>>>
>>> You'll probably need the sp, spdep and rgdal packages. You might
>also
>>> want to use the Raster package to extract the sampling data, or you
>can
>>> use spGRASS to tie the R and Grass together.
>>>
>>> See chapter 9 (1st ed) of Applied Spatial Data Analysis with R.
>>> http://www.asdar-book.org/
>>>
>>> Enjoy,
>>> Alex
>>>
>>> On 03/11/2014 09:18 AM, Tyler Smith wrote:
>>>> Hello,
>>>>
>>>> We're preparing a field sampling program, and would like to
>determine
>>>> a minimum distance between samples to reduce/eliminate spatial
>>>> autocorrelation. I think a good approach would be to calculate a
>>>> mantel correlogram, and use the range of the correlogram as our
>>>> minimum sampling distance.
>>>>
>>>> * Questions
>>>>
>>>> 1) is this a reasonable approach
>>>> 2) if so, how best to do this?
>>>>
>>>> * Details
>>>> We have a vector map with the point coordinates of several hundred
>>>> potential sampling sites, and ~ 10 raster layers with appropriate
>data
>>>> to test for spatial autocorrelation (WORLDCLIM, soils). I could do
>>>> something like the following, but I'm not sure if there's a simpler
>or
>>>> more appropriate approach:
>>>>
>>>> 1) extract the raster data for each point
>>>> 2) save the data to csv; import into R
>>>> 3) calculate the spatial distances between points, after projecting
>>>> the lat-long data into an appropriate scale (?)
>>>> 4) calculate the climate distance using the WORLDCLIM data
>>>> 5) use the 'mgram' function in the 'ecodist' package to calculate
>the
>>>> actual correlogram between the spatial distance and climate
>distance
>>>>
>>>> Any suggestions on the approach or the methods would be welcome!
>>>>
>>>> Thanks,
>>>>
>>>> Tyler
>
>>
>
>_______________________________________________
>grass-user mailing list
>[hidden email]
>http://lists.osgeo.org/mailman/listinfo/grass-user

_______________________________________________
grass-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/grass-user
Reply | Threaded
Open this post in threaded view
|

Re: mantel correlogram

Thomas Adams-2
In reply to this post by Alex Mandel-2
I agree, there -IS- "No need to pass csv around" because using the R spgrass6 package, one can read/write GRASS vector and raster files directly from R, so there are no intermediate files. I do this "all the time" — incredibly powerful using GRASS & R together.

Tom


On Tue, Mar 11, 2014 at 1:08 PM, Alex Mandel <[hidden email]> wrote:
You are right, I didn't read it that closely 1st time around. My point
was that all of it can be done in R, and there are geospatial specific
packages that have all the tests one might want. The bare minimum
interaction is via rgdal or spgrass to pull data over from existing
GRASS data sets. If the data isn't already in GRASS then rgdal one can
read the original files directly. No need to pass csv around. Of course
if it is in GRASS then you should have it all in the same projection
already anyways if you put it all into the same mapset/location.

The other book likely to have exactly what you want (field sampling
design) is Ch 5.
http://www.amazon.com/Spatial-Analysis-Ecology-Agriculture-Using/dp/1439819130/ref=la_B001K6MGR8_1_1?s=books&ie=UTF8&qid=1394557436&sr=1-1

Enjoy,
Alex

On 03/11/2014 09:58 AM, Thomas Adams wrote:
> Alex,
>
> I believe Tyler does plan on using R for the statistical analyses, but
> using GRASS GIS in combination with R is the easiest path, I think.
>
> Tom
>
> On Tuesday, March 11, 2014, Alex Mandel <[hidden email]> wrote:
>
>> Use R. It includes Moran's I and Geary's C tests for
>> spatial-autocorrelation. Look like it has  mantel too.
>>
>> You'll probably need the sp, spdep and rgdal packages. You might also
>> want to use the Raster package to extract the sampling data, or you can
>> use spGRASS to tie the R and Grass together.
>>
>> See chapter 9 (1st ed) of Applied Spatial Data Analysis with R.
>> http://www.asdar-book.org/
>>
>> Enjoy,
>> Alex
>>
>> On 03/11/2014 09:18 AM, Tyler Smith wrote:
>>> Hello,
>>>
>>> We're preparing a field sampling program, and would like to determine
>>> a minimum distance between samples to reduce/eliminate spatial
>>> autocorrelation. I think a good approach would be to calculate a
>>> mantel correlogram, and use the range of the correlogram as our
>>> minimum sampling distance.
>>>
>>> * Questions
>>>
>>> 1) is this a reasonable approach
>>> 2) if so, how best to do this?
>>>
>>> * Details
>>> We have a vector map with the point coordinates of several hundred
>>> potential sampling sites, and ~ 10 raster layers with appropriate data
>>> to test for spatial autocorrelation (WORLDCLIM, soils). I could do
>>> something like the following, but I'm not sure if there's a simpler or
>>> more appropriate approach:
>>>
>>> 1) extract the raster data for each point
>>> 2) save the data to csv; import into R
>>> 3) calculate the spatial distances between points, after projecting
>>> the lat-long data into an appropriate scale (?)
>>> 4) calculate the climate distance using the WORLDCLIM data
>>> 5) use the 'mgram' function in the 'ecodist' package to calculate the
>>> actual correlogram between the spatial distance and climate distance
>>>
>>> Any suggestions on the approach or the methods would be welcome!
>>>
>>> Thanks,
>>>
>>> Tyler

>




_______________________________________________
grass-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/grass-user
Reply | Threaded
Open this post in threaded view
|

Re: mantel correlogram

Markus Neteler
In reply to this post by Tyler Smith-2
On Tue, Mar 11, 2014 at 7:43 PM, Tyler Smith <[hidden email]> wrote:
> Thanks for your suggestions. It looks like the R Borg is continuing to assimilate procedures that once required specialty software. Time to learn some new packages!

Please consider to eventually add some short workflow here:

http://grasswiki.osgeo.org/wiki/R_statistics

More use cases are needed in this page in my view, would be great to
have yours then.

Best
Markus
_______________________________________________
grass-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/grass-user
Reply | Threaded
Open this post in threaded view
|

Re: mantel correlogram

Thomas Adams-2
Markus,

I can do that -- how do I make example datasets of mine available (which I would prefer, instead of using SPEARFISH or one of the other common datastes)?

Tom


On Sat, Mar 15, 2014 at 9:48 AM, Markus Neteler <[hidden email]> wrote:
On Tue, Mar 11, 2014 at 7:43 PM, Tyler Smith <[hidden email]> wrote:
> Thanks for your suggestions. It looks like the R Borg is continuing to assimilate procedures that once required specialty software. Time to learn some new packages!

Please consider to eventually add some short workflow here:

http://grasswiki.osgeo.org/wiki/R_statistics

More use cases are needed in this page in my view, would be great to
have yours then.

Best
Markus





_______________________________________________
grass-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/grass-user
Reply | Threaded
Open this post in threaded view
|

Re: mantel correlogram

Markus Neteler
Tom,

On Sat, Mar 15, 2014 at 3:10 PM, Thomas Adams <[hidden email]> wrote:
> Markus,
>
> I can do that -- how do I make example datasets of mine available (which I
> would prefer, instead of using SPEARFISH or one of the other common
> datastes)?

if not too big and if having an open data license we could host them
on an OSGeo server (incl. the GRASS server).

Markus
_______________________________________________
grass-user mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/grass-user