Combining algorithms based on robust estimation techniques and co-integrating restrictions

✍ Scribed by J. Hallman; M. Kamstra

• Publisher: John Wiley and Sons
• Year: 1989
• Tongue: English
• Weight: 573 KB
• Volume: 8
• Category: Article
• ISSN: 0277-6693
• DOI: 10.1002/for.3980080305

✦ Synopsis

This paper will motivate alternative combining schemes, provide a statistic for comparing alternative combinations out-of-sample and provide an example demonstrating these techniques. The evidence suggests the proposed procedures are likely to d o no worse than other approaches and promise to do better under circumstances commonly encountered with economic data: integrated series and contaminated data.

KEY WORDS Combining forecasts Robust estimation

Co-integration Encompassing 'For a detailed deicriprion of the algorithms performing these robust estimations see Coleman et al. (1980) and Kuppert and Carroll (1980).

'A number of the robust procedures not reporred here also had smaller RMSFE than both the individual forecasts. 'Within sample, an OLS combination trivially encornpasse5 its components. since OLS residuals are orthogonal to independent variables.