Project Description and Background The Edinburgh Solicitors Property Centre (http://www.espc.co.uk) has quarterly time series for local house prices dating back to 1993. In terms of time series modelling twelve years of quarterly data is a relatively short series, particularly since degrees of freedom are likely to be reduced by a considerable number of potential explanatory variables and lagged variations of these. One expedient might be to forecast Scottish house prices in general and use such prices as an explanatory variable for Edinburgh prices. In this way the Scottish prices would act as a conduit for other variables. The Nationwide Building Society has regional quarterly series for house prices (http://www.nationwide.co.uk/hpi/historical.htm) from 1973. The Office of National Statistics provides free access to a large number of potential explanatory time series (http://www.statistics.gov.uk/statbase/tsdintro.asp). What is involved? The objective of this project would be to develop suitable interacting models for predicting Scottish and Edinburgh house prices. The primary basis of assessment would not be the production of an excellent model, but demonstration of a correct and insightful methodology in producing such a model. This last sentence bears reading several times; lucky number crunching will avail little - regardless of the result - unless accompanied by lucid and insightful commentary. However, neither will clever commentary avail much if it is not evident that techniques have been correctly executed. Examiners should be fully persuaded of mastery of both techniques and their underlying principles. The project would involve collecting data, learning SPSS and EXCEL, examining a variety of forecasting models. These models would include Winters exponential smoothing, various types of regression model, seasonal ARIMA models, and transfer function models. The project report would include a consideration of the various models examined in terms of the statistics pertaining to each. Skills Needed The student may come to the project with suitable training in statistical forecasting. Otherwise, in order to complete this project the student would have to teach himself the material in the first 8 Chapters of Forecasting Methods and Applications (1998) Makridakis, Wheelwright and Hyndman (http://go.to/forecasting). Altogether this amounts to mastering a full university course as a preparation to conducting the project analysis by self-instruction, and accordingly, the amount of analysis required of the project would be more modest than would be the case were such preparation not required. The student would agree on a schedule of progress to be made on mastering the statistical and forecasting material, and progress would be checked on the basis of periodic assignments.