The thesis is dealing with an analysis of voting behavior in
the Liberec Region in 1992-2010 using techniques of exploratory
spatial data analysis (ESDA) at the level of municipalities.
The research introduces methods, which are relatively new, and
in the Czech political science, still not very used. These
methods also seek to contribute to, at least partially, solve
the problems associated with the analysis of spatial data.
Specifically, we use: a) Moran’s I statistic to describe global
extent of spatial clustering (spatial autocorrelation) of the
analyzed phenomena; b) local indicators of spatial association
(LISA) to map areas with different character of spatial
autocorrelation; c) geographically weighted regression (GWR) to
look for a connection between electoral support for the
political parties and the basic characteristics of
municipalities and theirs populations. Factors behind the
voters decisions are being sought by a combination of
independent variables that represent two basic approaches that
are used in electoral geography to explain spatial variability
of election results – composite and contextual approach.