P E N U M B R A

The work summarized in the poster was titled as "Exploratory spatial analysis of theft as a function of land use and socio-demographic data". The emphasis here was more on methodology rather than the outputs or the specific case used. In a brief, we used OLS (classic) regressions to investigate how large the impact of socio-demographic and land-use indicators is on the ratio of theft in Tallinn. The independent variables were derived from the two prominent theories in criminology, routine activities and social disorganizations, and included the followings:

number of pubs, number of bus stops, number of house owners, male population, population of foreigners, proportion of unemployed.

The overall procedure initiated with normalizing the data for the independent variable (theft), running univariate regressions for theft and each of predictors and checking p-value, goodness-of-fit and the diagram in search of a significant relationship. Since we did not observe any remarkable relation, proceeded then with a visual ESDA test: we used box-plots and maps to select and visually compare top positive residuals of y and x variables and came up with the same result: no significant relation between theft and socio-demographic and land-use factors.

The literature, however, suggests some relation between the two. The source of error in our experiment could have been among these:

- We should have normalized data in another way than using logs of the values.
- One multivariate regression should have been performed instead of a set of univariate models. This considers cumulative impact of the factors and will thus better account for the overall impact of socio-demographic and land-use parameters on theft.
- Even though the results of OLS regression were not satisfactory, we should have proceeded and also performed spatial lag and spatial error regressions hoping that they may better fit the data.

This was however a basic and limited spatial analysis study performed in a short time and has much room for further development and methodological modifications.