search query: @indexterm Business forecasting / total: 147
reference: 32 / 147
| Author: | Smith, J. O. |
| Title: | Non-linear state space models with partially specified distributions on states |
| Journal: | Journal of Forecasting
1990 : MAR-APR, VOL. 9:2, p.137=149 |
| Index terms: | BAYESIAN STATISTICS FORECASTING DYNAMIC MODELS BUSINESS FORECASTING |
| Language: | eng |
| Abstract: | A class of non-normal/non-linear state space models is defined. These models are useful for business forecasting which is rigourous, easily interpretable , exact and computationally tractable. Through the discussion of several examples, it is shown why models like the power steady models are attractive. The simplicity of their one-step-ahead forecast distributions is emphasized. Furthermore they can be justified through sets of properties that any gradually evolving process may be expected to satisfy. In most business applications the types of partially specified models have many advantages, however, they are not suitable for all purposes, and care must be taken in using them. |
SCIMA