Hi there. I'm trying to help a client out -- she's using hotpost analysis (Gi*) to evaluate the prevalance of (for confidentiality purposes, I'll call it "Attribute A", across various districts over a geographic area. She's found a few clear hotspots that warrant further analysis. Her research team has suggested that, in addition to this, she also run a z-test (or t-test) to compare the values in the hotspot analysis with those outside of it as a further test of significance. I'm not a statistician, but does this make sense? I'm skeptical for a couple of reasons:
1)these latter tests assume normal distribution (which wouldn't hold for this dataset), and 2) there's already significance testing built into the Gi* calculation. Is there a reason to further compare the "Attribute A" levels in hotspot regions with non hotspot regions? I know Gi* won't answer questions about underlying processes and further evaluation is needed, but it seems like the way to go is to evaluate other attributes as well in addition to "Attribute A". I've also thought that evaluating the Gi* over various distance bands (perhaps in conjunction with a Ripley's K) might be interesting to look at clustering over spatial scales, but I'm not sure any of this would really tell us anything new. Thoughts?
Thanks!
