An application of FDSM to seasonal adjustment.

This dissertation studies a new seasonal adjustment technique within the framework of a general linear model. The research also presents comparisons to alternative seasonal methods, namely, X11, X11ARIMA, and BAYSEA, in examining futures price variations. Different approaches for analyzing seasonal variations are briefly reviewed, and a new approach to seasonal adjustment, the fractional differencing seasonal model, is examined in the dissertation. The best linear unbiased predictor for the variable being analyzed is developed when the disturbance components of future values of the variable are correlated with its current and past disturbances; and a FORTRAN program is written for the calculation of the maximum likelihood estimates for mean, seasonal factors, variance, and the maximized log likelihood of given time series. Several criteria for assessing the performance of methods of seasonal adjustment are reviewed, and they are applied in analyzing the volatility of futures commodity prices. Six commodities are investigated, namely, soybeans, T-bills, copper, cotton, pork bellies, and frozen orange juice. Emphasis is placed on the seasonal pattern of futures price variations for these commodities. By analyzing a large collection of many prices of futures contracts for these commodities using several seasonal adjustment methods, including the fractional differencing seasonal model, the data analysis presented in this dissertation attempts to provide a better underst and ing of seasonal and maturity effects on futures price volatility.Ph.D.FinanceStatisticsBusiness administrationUniversity of Michiganhttp://deepblue.lib.umich.edu/bitstream/2027.42/161784/1/8812871.pd