Robust and Non-Robust Analysis of Semivariogram Isotropic in Crime Data by Changing Sill, Case Study: Bandung’s Theft Data
Keywords:
Cressie-Hawkins, Crime, Dowd, Loss value, Matheron, Range, Sill, TheftAbstract
Research in criminality is sufficiently developed but mathematical statistics analysis has little role in this field. In
Indonesia, this research is mostly done in descriptive statistics and simple modeling. The population development has an effect on social and economy. Consequently, the crime rate increases with the compliance of people's living needs. In this research, we focus on analyzing criminal loss caused by theft. The loss is modeled by an isotropic semivariogram model. Here, we consider the non-robust Matheron model and the robust Cressie-Hawkins and Dowd models to analyze semivariogram of the crime. The best model is determined from the variance of loss and statistics of experimental semivariogram such as mean, first quartile,
median, and third quartile. Data with enough high loss has
candidate of sill that is first and third quartile of experimental semivariogram. We apply the analysis to Bandung’s theft data and corresponding model is exponential with a range of 4.45 kilometers. Through this model, we can predict theft having significant losses at the range of 4.45 kilometers. This information can be a recommendation for the police to raise awareness for locations around 4.45 kilometers from the location of theft.