CMStatistics 2017: Start Registration
View Submission - CMStatistics
Title: Intervention variables and stochastic trend in state space models Authors:  Elisa Jorge-Gonzalez - Universidad de La Laguna (Spain) [presenting]
Enrique Gonzalez-Davila - Universidad de La Laguna (Spain)
Raquel Martin-Rivero - Universidad de La Laguna (Spain)
Domingo Lorenzo-Diaz - Universidad de La Laguna (Spain)
Abstract: Time series are often affected by external phenomena that cause a temporary or permanent change in the expected behavior of the same. There are different ways in which this type of phenomena or interventions influence the development of the series, depending on the duration and the form of the impact caused. The most common interventions are the step effect, a sudden change with permanent duration, and the impulse, a transient point change. These effects are usually incorporated into the models by including intervention variables that improved estimates, regardless of the method of analysis used in the study of the series. The use of state space models in the analysis of structural time series allows, among others, the specification of distributional conditions on the level and trend of the series. The main objective is to analyze the utility of state space models with stochastic trend and slope when there are intervention variables with different degrees of impulse. For this purpose, simulated data will be used, as well as an example of real data of airplane passengers arriving in the Canary Islands.The CajaCanarias Foundation Project, PROYECTUR 2015 TUR03, has supported this work.