Need to test to spatial autocorrelation?

Discussion created by dibishop on Aug 21, 2012
Latest reply on Aug 22, 2012 by jevans02
I'm trying to determine if testing for spatial autocorrelation is necessary and/or advisable with my small study sample. I'm working on the analysis for a study with an ecologic study design. A total of 10 census tracts in one city are included in the analysis (5 intervention/5 control group) where events over three time points are being examined.  X/Y coordinate events have been summed for each census tract at each time point, and then are totaled for each the intervention and control group. The outcome for the study is the total number of these events -- examining if there were significant changes over time in the intervention tracts compared to the control tracts.

Is testing for spatial autocorrelation necessary for this type of study design? If so, is the N=10 too small to reliably calculate any of the tests for autocorrelation?

Thank you in advance for any help/guidance!