A major consequence of the dependence in a spatial sample is that statistical inference will not be as efficient as for an
independent sample of the same size. In other words, the dependence leads to a loss of information.
1
Roughly speaking, and
everything else being the same, this will be reflected in larger, variances for estimates, lower significance levels in tests of hypotheses
and a poorer fit for models estimated with data from dependent samples, compared to independent samples of the same size. I will
refer to this aspect of spatial dependence in the rest of the paper as a nuisance. The loss in efficiency may be remedied by increasing
the sample size or by designing a sampling scheme that spaces observations such that their interaction is negligible. Alternatively, it
may be taken into account by means of specialized statistical methods. In this paper, I will focus on the latter. When spatial
dependence is considered to be a nuisance, one only wants to make sure that the interpretation of the results of a statistical analysis are
valid. One is thus not really interested in the source of the spatial association, i.e., in the form of the spatial interaction, the
characteristics of the spatial structure, or the shape of the spatial and/or social processes that led to the dependence. When the latter is
the main concern, I will use the term substantive spatial dependence instead.