topic Re: How do you choose the time interval and distance treshold to generate appropiate Spatial Weights matrix? in Spatial Statistics Questions

How do you choose the time interval and distance treshold to generate appropiate Spatial Weights matrix?

CharlotteDe_Grave — Fri, 05 Aug 2016 16:07:16 GMT

Hello everybody,

I have a set of data points corresponding to diseased trees on a particular date. I want to do a grouping analysis based on space and time through the generation of a spatial weights matrix. I wanted to know how to choose appropriate time intervals and treshold distances to generate the matrix. I tried some combinations of time intervals and distance tresholds but than, it wasn't possible to run the grouping analysis because "there were too many disconnected groups to solve this model". I was able to find a combination that allow the running of a grouping analysis but than, I don't know how to scientifically account for this choice of values.

Can anybody help me with this?

A big thanks in advance,

Charlotte

Re: How do you choose the time interval and distance treshold to generate appropiate Spatial Weights matrix?

DanPatterson_Retired — Fri, 05 Aug 2016 21:56:39 GMT

Charlotte... from here How Generate Spatial Weights Matrix works—Help | ArcGIS for Desktop

there are a few limited resources available,... but the gist in parts, is you are attempting to represent your problem in a spatial context. So you must have some sense or feeling as to what is going on. Simply looking a set of diseased trees in an area at a particular point in time is going to be a difficult problem to explain. You must have some way of articulating what you feel the process is and how you want. So what parameters did you use? or could use?

You probably know that the likely of disease spread by insects or other airborne vectors is going to be proximal based and euclidean distance (ie not road distance/manhatten etc). How far? that comes back to the vector... is it wind borne, insect etc. This is one area of statistics where you enter the area of not necessarily having a hard p-value statistics. Exploring other options within any of the spatial tools may reveal things that you hadn't considered.

Re: How do you choose the time interval and distance treshold to generate appropiate Spatial Weights matrix?

CharlotteDe_Grave — Mon, 08 Aug 2016 14:20:27 GMT

Hey Dan,

Thank you for your response but i am not sure I have understood you well.

I tried to put parameters that are logical in regard to what i want to study but then, the grouping analysis was not possible for the reasons I described above.

I am not sure how the spatial weights matrix is constructed and I don't know how to read this kind of files (.swm).

Re: How do you choose the time interval and distance treshold to generate appropiate Spatial Weights matrix?

DanPatterson_Retired — Mon, 08 Aug 2016 16:16:33 GMT

Your statement here

I tried some combinations of time intervals and distance tresholds but than, it wasn't possible to run the grouping analysis because "there were too many disconnected groups to solve this model

You tried... based upon the knowledge of the processes? or you just tried some combinations?

You will have to help us with what you are trying to do. The swm is needed for some tool options, but what are you going to use?

Modeling spatial relationships—Help | ArcGIS for Desktop

Re: How do you choose the time interval and distance treshold to generate appropiate Spatial Weights matrix?

CharlotteDe_Grave — Thu, 11 Aug 2016 12:52:39 GMT

I tried upon the little knowledge of the spread of the disease that we have.

I am studying Laurel wilt spread in avocado groves. I want to see if there are any trends in the spreading, based on time and space. I would like to answer the question if it is only spreading from avocado tree to avocado tree or if there is a continuous inlet from outside the avocado growing area.

That's why i wanted to do a spatial weights matrix and than a grouping analysis, to see if the data inside the growing zone is closely related.

I don't know if it is the best way to answer the question.