The project involves selecting, analyzing and writing a report on a univariate time series. The report is due in class on March 4, but may be submitted earlier.
Search for a time series that interests you (the project is meant to be fun). The web is an excellent resource for data. A list of sites providing data on various topics including economics, the environment, demography, health and finance data is available at http://www.stat.lsa.umich.edu/~ionides/531/ts-data.html . A wide variety of time series are also available from the Time Series Data Library, www-personal.buseco.monash.edu.au/~hyndman/TSDL/ . Google may also help. Please see me if you have problems finding an appropriate data set.
Look for a time series that is approximately stationary, or may be made so by detrending and/or a transformation. Try to have at least 100 equally spaced observations in the data set. You should not knowingly use the same data as anyone else in the class, though you should feel free to discuss your project.
The analysis should be done using R. You should not usually present your code. However you may comment on R commands, or other computational issues, of particular interest. It is expected that the report will contain a graph of the data, a sample a.c.f., an estimated spectral density, a fitted ARMA model, some residual analysis and discussion of the results.
The report should begin by presenting some detail concerning your data and where you obtained it. You should aim to use the data to address carefully at least one question, concerning the subject matter, that interests you. The report is to be no longer than 5 pages, including figures. Marks will be lost if it exceeds that length. Think of the midterm project as practice for the longer final project. The final project will consist of analyzing the relationship between two or more time series. The final project can be a continuation of your midterm project, so you might like to bear this in mind when selecting your problem.