Statistics 4540: Term Project
Written Report: Due no later than April 12, 2004
Presentation: April 7 or 8, 2004

For your project, you are required to find a time series dataset and conduct a thorough analysis, based on methods discussed in class and in your textbook. This includes using both time domain and frequency domain methods. I will give you some websites where lots of time series examples are available.

Your written project should be split up into different sections, which contain complete paragraphs and sentences. Your report can be typed or handwritten. I will also expect you to submit any relevant computer output, as described below. This does mean that I will not want to see things Splus did that you did not want it to do. In particular, please do not print your entire dataset and submit it with your report.

You will be graded on the clarity of presentation, statistical methodology, grammar, spelling-in short, all the things that will be important when you prepare these kinds of reports for a living as a paid statistician.

I will give another handout in a couple of weeks on what I'll expect for the presentation you will give on your project.

A general outline of the format is given below. Note that not all sections will be of equal length.

Format of Report

In preparing your project, break your report up into the following sections, and use these sections titles to aid in making your project clear and easy to follow:

• Introduction: Include a description of data, and what you will try to do in your analysis.

• Exploratory Analysis: Include things like plots, tables, summary statistics. Not all of these may be needed or applicable. Plots should include (but may not be limited to) a plot of raw data, ACF and PACF of raw data, plots from the frequency domain that we discuss, plots after any transformations due to nonstationarity, and so on.

• Analysis and Modelling: This will include discussion on the potential models for your time series, and parameter estimation for those models. Although we haven't discussed this much in class, you should include some inference (confidence intervals or hypothesis testing) for your model parameters. You can see some of the examples in the text (especially Chapter 2) for details. You should also do include any appropriate methods discussed in the frequency domain approach. When choosing a suitable ARIMA model for your data, be sure to explain your choice.

  For the final model you select, you should use some model diagnostics (residual plots,etc.) to see if your model assumptions are satisfied. You should also use your model to demonstrate its forecasting ability.

• Conclusions: Brief summary of findings.

• Appendix: Containing any computer output.

Finding Data

Here are some website that contain a large collection of time series data. The graduate students should also investigate these sites to try and find some suitable data to go with the paper they are studying.

• www.library.mun.ca/qeii/govdocs/govdocs.php#statistics

  and click on the link for CANSIM II. This site contains tons of socio-economic and demographic time series data.

  WARNING: CANSIM II is easiest to access if you're accessing the internet at MUN. Just click on the link and you get access. If you're trying to access it from home, you have to follow the directions that are given beside the CANSIM II link.

• www-personal.buseco.monash.edu.au/~hyndman/TSDL/

  This site contains about 500 time series from many different fields. This site also has some links to other time series collections and resources on the internet.

• www-unix.oit.umass.edu/~statdata/statdata/stat-time.html

  A small collection of data sets.

• dss.ucar.edu/catalogs/free.html

  A large collection of atmospheric and oceanographic datasets. However, a lot of the datasets are really big, or measure variables that may not be familiar to you.

You are welcome to find a time series from a different source than the ones listed above.

When selecting a time series, please make sure that it is interesting: it has a trend and/or a periodic (seasonal or otherwise) component, etc.