A. Iparragirre, I. Barrio, I. Arostegui
Prediction models are widely used in daily practice. Increasingly, experts of different research areas are collecting data using a complex sampling design. Although many researchers have claimed the importance of considering the design of the survey when fitting prediction models, in practice, often this information about the survey is not taken into account in the estimation process. The aim of this work was to develop logistic prediction models based on complex survey data considering the sampling weights either in the estimation or the validation process. We use the fitted model to make predictions for the individuals in the entire finite population and we propose to consider the sampling weights to select the optimal cut point to classify these population individuals as event or non-event. In particular, we have applied this methodology to a complex survey, which was designed, collected and provided by Eustat (Euskal Estatistika Erakundea - Instituto Vasco de Estadística).
Palabras clave / Keywords: complex survey, prediction, validation
Programado
Pósteres I
30 de mayo de 2018 15:30
Zona EXPO
