kkober

sign results from hot spot analysis

Discussion created by kkober on Apr 15, 2011
Hi all,

I am working with the hotspot tool in ArcGIs and have a rather mathematical question. The notes state that one can use the z-score to identify significant results. It is worth noting that this only works when the values have a normal distribution, though!

Since my data is not normally distributed I consulted with a statistician who advised me to simply choose the top 1% (or 5%, whatever I would have chosen as a critical value in a test) of my results. He stated this would do something similar to a significance test, but would not be based on the assumption of an underlying frequency distribution of the data.

So, how do I choose the top 1%? I've got two possible methods:
(a) top 1% values: just count the values I've got, calculate what 1% of this number is and choose the highest values until the number reaches the required number for 1% (e.g. I've got 1000 values and would then need to choose the highest 10 of these to get the top 1% values)
(b) top 1% of "population": sum up all values. Identify what 1% of this sum would be. Choose then the values by adding them(starting with the largest ones) until the required sum is reached (this makes sense when one tries to get 1% of a population of animals distributed over a map).

Has anybody an idea which of these methods resembles best what a statistical test is doing?
Sorry about this geeky question! And thanks for the help!!!

Kerstin

Outcomes