How can a model account for unobserved site characteristics?
In our models, we can include an extra variable that represents everything at that site that we have not measured or observed. When we estimate that model, we have therefore accounted for the variation in the modeled variable explained by those site specific characteristics that went unobserved. Using this approach one cannot then determine what contribution individual unobserved characteristics contributed but can see the effect of all the characteristics combined. Certainly this is inferior to having explicit information on site characteristics, but it is an acceptable method to capture the effect of these characteristics without the expense of gathering the specific information.
