Club of Statistical Journals Review

El Club of Statistical Journals Review es un espacio académico dedicado a la presentación y discusión de artículos de investigación en estadística, sin restricción de época, enfoque o temática. Los artículos revisados pueden ser históricos, prácticos o teóricos, permitiendo así una visión amplia de la disciplina.

El objetivo es generar un espacio colaborativo donde el estudio individual se transforma en aprendizaje colectivo.
2025-09-08
14:50hrs.
Fabián Gómez. Pontificia Universidad Católica de Chile
The discriminative functional mixture model for a comparative analysis of bike sharing systems
Por confirmar
Abstract:
Functional data analysis is a branch of statistics that, instead of dealing with discrete data as in classical statistical analysis, focuses on continuous data that are modeled as functions. One area that has been explored in the last decade concerns the clustering of functional data with the aim of determining clear patterns or behaviors in a high-dimensional context. This work studies the implementation of a functional mixture discriminant model (FDM) with parameters estimated through a functional extension of the EM (FEM) algorithm, with the objective of performing clustering for subsequent analysis. First, a simulation study is carried out to evaluate the model’s ability to correctly classify observations in contexts where the data are smooth and also when they are not, achieving good classifications in both cases. Then, an application is performed using data of bike sharing systems, comparing models fitted with different numbers of clusters and selecting the best one using AIC and BIC criteria.