Seminari – Dott. Haedo
19 marzo 2009
Luogo: Alma Mater Studiorum Università di Bologna Dipartimento di Scienze Statistiche “Paolo Fortunati” via delle Belle Arti 41 Bologna Aula 3 ore 15
Il Dott. Christian Haedo Coordinator of Research Center of Bologna University at Buenos Aires Poject, coordinator of Regional SMEs Observatories in Latin America, Coordinator of Master in Market Research and Data Mining of Bologna University at Buenos Aires, ha tenuto un seminario dal titolo:
Non parametric techniques for the identification of specialized agglomerations in continuous space. (Farall, Haedo, and Mouchart)
The spatial process approach is used to analyze geocodified data, and aims at assessing the power of attraction from a local space perspective.
From the point of view of a non-homogeneous Poisson process, location points are randomly distributed, and disjoint area counts are mutually independent, each based on Poisson’s distribution according to which the intensity parameter forms a finite measure of the reference space, in this case a bi-dimensional space.
This measure may be interpreted as the representation of the differentiated power of attraction of the space and, in the case of a specific area, the expected value of the number of locations in such area.
The availability of geocodified data unleashes the need to quantify the specialization level of a certain phenomena’s in a particular point of the space.
With this purpose, we present the methodology that uses kernel density estimators as a key tool to define a local specialization measurement for a point x as an extension of the well-known local quotient measurement to the continuous space.
For the identification of specialized agglomerations in continuous space, we base in the identification of statistical significance or significant feature of the specialization level based on the method of bootstrap hypothesis testing proposed by Efron and Tibshirani (1993) to approximate the distribution of the local quotient under the non-specialization hypothesis.
The study cases of this work first analyze simulated data and next use geocodified data of firms of manufacture sector of Buenos Aires City at four digits of ISIC Rev.3.