I am new to ArcGIS and am looking to find outliers (high and low) in my data. The variable I am using is a disease severity variable ranging from 0 to 60 (i.e. my analysis field). I am trying to find outliers in this variable. However, I recently figured that each time I run the Optimized outlier analysis or the cluster and outlier analysis, there is a good proportion of observations with changed COtypes/p-values, or that I am finding different outlier observations on each run. Although I see that a good proportion are the same when repeated, I am concerned with this change and don't understand why this is happening. I would really appreciate it if anyone could explain why this happens with outlier analysis (Local Moran’s I?) and can I trust that the outliers produced, although different in each run are actually outliers that I can use for my study? Thank you very much for your kind responses.
Different results can happen for a couple different reasons.
- Parallel processing can affect precision of the analysis which can create slightly different results from run to run.
- Permutations will also create slightly different results. You can fix this in one of 2 ways. You can set the permutations parameter to 0 or you can use the Random Number Generator Environment. If you set the random seed to anything other than 0 (let's say 2 and then use the seed 2 every time), you should receive the same results every run.
Hope this was helpful. Let me know if you have any other questions.