Hi Bipin,

most of spatial statistics techniques are successfully used in fields of spatial epidemiology, and health geography. So you can find useful resources in this fields. However other applications may be easily derived based on this information. I would recommend you e.g.

Pfeiffer, D., Stevenson, M.D., Robinson, T., Rogers, D., 2008. Spatial analysis in epidemiology. Oxford University Press.

Waller, L.A., Gotway, C.A., 2004. Applied Spatial Statistics for Public Health Data. John Wiley & Sons.

Bivand, R.S., Pebesma, E.J., Gómez-Rubio, V., 2008. Applied Spatial Data Analysis with R. Springer New York, New York, NY. doi:10.1007/978-0-387-78171-6

where you can find well-structured chapters about an evaluation of spatial pattern and/or spatial clustering.

Regards

Lukas

http://gis.humboldt.edu/OLM/r/Spatial%20Analysis%20With%20R.pdf

Article Spatial Analysis in Epidemiology

Dear Lukas,

Thank your very much for your reply. I will go though  these literature.

Bipin

As outlined in the chapters on disease clustering tests in Waller and Gotway (2004), I think the most important part of any comparison of methods is to determine what question each method addresses.  Moran's I gives a global summary of whether values show positive spatial correlation (p-value for the entire data set), SaTScan identifies the most unusual clusters (p-value and where they are).  Local versions of Moran's I (and other general indices) give some insight on local clusters but in a different way that SaTScan. 

Dear Lance,

Thank you very much for your reply.  As you said Moran's I is used to assess the global auto correlation or global cluster however  there are other spatial statistics such as Local Moran's I and Local G statistics which measures the local clusters. Now SaTScan and Local Moran's I and G Statistics appears similar. I have read somewhere that spatial statistics  use location (x,y) to measure cluster but SatScan use population base (threshold) for the same.   Now my question is SaTscan  or Local Moran or G Statistics is better to explorer local cluster in the data set. 

Local Moran's I, G statistics, and SaTScan answer slightly different variants of the same question.  It is not really a question of which is better in terms of power, etc, as they respond to slightly different signals in the data.  SaTScan finds the most likely clusters and determines how unusual that appears among the possible most likely clusters under H0: No clusters.  The LISAs identify which areas are most similar or different to their neighbors where H0: local independence.  Both can be used with the locations alone or adjusted for background populations by converting to local rates. 

This is clearer when you think about what you get.  SaTScan gives the location, size, and p-value for the most likely clusters (maybe a few on your map).  These are the most suspicious areas (but may not be significantly suspicious).

LISAs give you a full map with each region (or location) colored by how similar its value is to the values of its neighbors.  All regions (or locations) get a value and they are colored by being high or low.  Significance is based on each region (each region gets a p-value compared to what you would expect for that region under H0: independence) so there are multiple comparison issues.  SaTScan gives p-values comparing the score from the most likely cluster to the score of most likely clusters under H0: no clustering, regardless of where they may occur under the null.

So it isn't really a question of which is best but a question of what are you looking for.

The description in the Waller and Gotway chapters outlines this in more detail.   Also the book by Tango on tests for disease clustering provides a nice description with comparison and contrast.  I hope this helps.

Thank you Lance for your reply and well explanation.

Excellent discussion here on Moran's I and SaTScan. Thank you.

To my experience, it is better focus on the objective of the authors for the selection of either traditional or new spatial statistics. I liked all the discussions made before. They are interesting since they are forwarded based on people's life experience.

I am happy and share with responses given by Lance Allyn Waller . All the points that a person is asking for are mentioned there.