I'd like to attempt a regression analysis that can help me understand the relationship between septic tanks, landuse patterns, and population (independent variables) and nitrogen concentration (dependent variable) in an estuarine waterbody. I've read in numerous places that spatial dependence between variables is a fundamental problem when applying statistics to spatial data. From what I understand, in order to apply ordinary least squares regression, the data cannot exhibit spatial dependence (or is it the residuals of the regression model that cannot exhibit spatial dependence, correct me if I'm wrong)? I expect to see higher concentrations of nitrogen in the estuary clustered around nitrogen sources such as septic tanks, high population density, and certain landuses.
Does a Moran's I test identify spatial dependence between dependent and independent variables?
Is spatial autocorrelation simply a measure of spatial dependence/independence?
And if Moran's I were to show significant clustering between dependent and independent variables, would I not be able to continue with a regression analysis (or, again, is this only for the residuals of a regression model)?
Thank you so much for anyone who can help me better make sense of this.