Classification trees for poverty mapping : a thesis presented in partial fulfilment of the requirements for the degree of Master in Applied Statistics at Massey University, Palmerston North, New Zealand
Measuring differences in poverty levels within a country is important for aid allocation.
Small area estimates of poverty incidence can be found by combining census
and survey data. The usual method uses multiple regression, but an intuitive alternative
is to build a classification tree for classifying households as poor or non-poor.
This research presents some preliminary results using this method, and compares
them to the traditional regression method.