I need to test for spatial autocorrelation in my data. In R, I used the Moran.I function from the ape package, and the moran.test function from the spdep package, but got strickingly different results. I think this may be due do the weights employed to calculate Moran's I: While in ape your weights are given by your distance matrix, in spdep you must specify spatial weights from a neighbors (nb) object, choosing a given style (binary, standardised, ect). Could anyone please clarify this and perhaps suggest which package is best?
Different weights will give different results - think of a chessboard - bishop's case weights will give positive autocorrelation lots of black- black joins and lots of white-white joins. A rook (castle) weighting system would show negative autocorrelation lots of black - white joins and a Queens's case would show no autocorrelation. Same underlying pattern on the board but very different answers. This is a good thing as you can set the weights to evaluate different theoretical schemes.
In this classic paper Peter Haggett evaluated 7 alternative weights structure and showed that an epidemic was dominated by hierarchical spread at the beginning and local contagion later
Peter Haggett (1976) Hybridizing Alternative Models of an Epidemic Diffusion Process, Economic Geography, Vol. 52, No. 2, Human Health Problems: Spatial Perspectives (Apr., 1976), pp. 136-146
You may be also interested in this guide by a colleague - free to download
https://www.researchgate.net/publication/258151270_An_Introduction_to_Mapping_and_Spatial_Modelling_in_R
Book An Introduction to Mapping and Spatial Modelling in R
Awesome example Kelvyn Jones, I'll use it on my lectures;)
Thank you very much for your reply and the guide.
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India