Using ARIMA models to predict prison populations

In this study a time-series model for predicting Louisiana's prison population was developed using the iterative Box-Jenkins modeling methodologyidentification, estimation, and diagnostic checking. The time-series forecasts were contrasted with results of regression models and an exponential sm...

Descripción completa

Guardado en:  
Detalles Bibliográficos
Autor principal: Lin, Bin-Shan (Autor)
Otros Autores: MacKenzie, Doris Layton ; Gulledge, Thomas R.
Tipo de documento: Electrónico Artículo
Lenguaje:Inglés
Publicado: 1986
En: Journal of quantitative criminology
Año: 1986, Volumen: 2, Número: 3, Páginas: 251-264
Acceso en línea: Volltext (lizenzpflichtig)
Volltext (lizenzpflichtig)
Journals Online & Print:
Gargar...
Verificar disponibilidad: HBZ Gateway
Palabras clave:
Descripción
Sumario:In this study a time-series model for predicting Louisiana's prison population was developed using the iterative Box-Jenkins modeling methodologyidentification, estimation, and diagnostic checking. The time-series forecasts were contrasted with results of regression models and an exponential smoothing model. The results indicate that the time-series model is the superior model as indicated by the usual measures of predictive accuracy. When compared with actual data the predictions appeared sufficiently adequate to meet the needs of the correctional system for short-term planning.
ISSN:1573-7799
DOI:10.1007/BF01066529