how to interpret regressions with spatial autocorrelation?

Hi guys,

I am currently doing a GIS and statistical project on whether there is an association between access to green space and deprivation at the neighbourhood level (LSOA, MSOA). I performed a linear regression and geographically weighted regression on both scales and got no significant results indicating no relationship which was unexpected. I then performed a Morans I which showed the deprivation in areas was highly spatially autocorrelated (0.67,p=0.001). In my discussion does it make sense to say that maybe the spatial autocorrelation of the deprivation data hampered my ability to distinguish an independent association which is why i didnt find one?

Basically I am just trying to find out statistically why I didn't get the expected results for my discussion. Any help would be greatly appreciated :)


TS Contributor
Besides calculating spatial autoc. on your predictors, have you tried to assess the autocorr. among the model residuals?
IN the literature I have read, they usually calculate autocorr. among residuals. It must be also said, however, that the sampling design should be performed in a preliminary stage of the analysis in order to alleviate the spatial autocorr. that may exists, or spatial autoregressive modelling strategies can be performed.

Can you elaborate a little more about the type of data you have?