A negative spatial correlation natural example is the way passengers take seats in public transport. They tend to try to sit away from each other with even spacing.
I'm working on my final paper about Spatial Econometrics and your explanation helped me a lot! But I think it's worth explaining the contiguity criterion: I'm used to using the Queen criterion instead of the Rook, and it confused me a lot.
Great video! but you didn't answer the initial question.. what is the Moran's I for the 2016 election? My home city of NYC has a negative Moran's I - the poorest county in New York State (Bronx) borders on the richest (Manhattan). Every town in Queens houses a different ethnic group.
Your video is very helpful. Woud you mind translating the video and uploading it to a Chinese video website? Because we can’t use RU-vid in China, I hope my friends can learn from it too.
A potential example of negative spatial correlation - the location of retail stores? If there is a retail store from a chain that isn't highly correlated, the surrounding area probably has a lower rate of retail stores?
thank you so much for this video! you're a fantastic teacher. I do have one question: is it possible to end up with a positive, but small value for Moran's I that is statistically significant? if so, is there a difference in how you would interpret a larger positive value versus a smaller positive value if both are statistically significant?
How would you recommend computing Morans I for Vectors with values from [0,1]? My first idea was to just subtract x_i with x_mean and take the length of that vector. The problem here is, that its not possible to get a negative value for the length, which leads into getting bad value to interpret. Second idea, is to just compute Morans I for every entry in our vector independent from another and taking the mean of these. Is there any better way?
Unless I don't know what borders are, I believe the I for the 2016 election is about 0.317, so pretty strongly physically correlated. This is using the naive weighting average above which is like the L_0 norm or something.
I live in FL and Moran's I would be about 0 (that is equal democrats and republicans). How ever, republicans cluster in the center of the state versus the coastline. Micro vs. macro have different results.
The answer for the future: Point 1 has 2 direct neighbours: 2, 5 Point 2 has 3 direct neighbours: 1,3,6 Point 3 has 3 direct neighbours: 2,7,4 Point 4 has 2 direct neighbours: 3,8 => 10 neighbours because of the symmetry: 2*10=20