New concepts in deterministic growth curve forecasting
作者:
J. Berny,
期刊:
Journal of Applied Statistics
(Taylor Available online 1989)
卷期:
Volume 16,
issue 1
页码: 95-120
ISSN:0266-4763
年代: 1989
DOI:10.1080/02664768900000011
出版商: Carfax Publishing Company
数据来源: Taylor
摘要:
Attempts to better the Performance of classical growth curve functions have met with limited success. Construction industry projects highlighted the need to improve deterministic models rather than the stochastic methodologies which are nearly always based on the former. New concepts (changed for the first time since 1825) are formulated and used to generate multi-component deterministic models. Six highly diverse case studies, of which three are presented, were used to test one model and its autocorrelation form. Trial forecast standard errors showed a drop of 50% when compared to classical and stochastic models. Among the by-products of this work are uses of normalisation, scaling and a simple statistical procedure to estimate linear constants. A different consequence of the new concepts thew light on the problem of predicting a consumption process in marketing. The major implications of this research show the import of the new concepts and diversification of the fields of study on deterministic modelling; and also the need to reappraise the functional interface with many of the underlying processes of growth.
点击下载:
PDF (1498KB)
返 回