(Чанг Вэй-Te. Анализ случайных эффектов регрессионной модели при предсказании срока годности артиллерийских порохов).Master of science Dissertation. Naval postgraduate school. 1995. - 88 p.Most gun propellant is stored at depots for a long time before it is used. While being stored, the quality of the gun propellant may deteriorate and become unstable. In an attempt to avoid disaster due to use of unstable gun propellant, accurate prediction of the safe shelf-life of gun propellant is necessary. The shelf-life estimation methods used currently for a group of similar gun propellant lots are based on a fixed effects regression model. This does not take into consideration the fact that samples from the same lot are more similar than samples between lots. To capitalize on this lot-to-lot variation when estimating the shelf-life, first, a random effects regression model is developed. Secondly, a combined mixed effects model is estimated. The estimated model is then used to predict not only the shelf-life of a group of similar lots but also that of each individual lot of 5754 NACO gun propellant stockpile. The results indicate that, first, the claimed shelf-life is not adequate and requires amendment. Next, the group shelf-life estimated can be relatively conservative compared to the individual shelf-lives. In view of potential opportunity loss due to safe individual lots being discarded, use of individual shelf-life is recommended.Contents:
Introduction.
Fixed effects regression model.
Random effects regression model.
Conclusions and recommendations.
