MTH6139 Time Series School of Mathematical Sciences

 General Information   A Time Series is a collection of observations, usually taken in time. For example: values of consumption in the UK over several years, unemployment percentage in the UK over several years, surface air temperature change for the globe, arrival phases of an earthquake, IBM common stock closing prices, chemical process temperature readings. As in many data, here too, some uncertainty of observations may be present. If it can be considered as random, then the time series is viewed as a random variable. A statistical analysis of the past data can reveal various features of the phenomenon of interest, for example whether there is any trend (i.e., increasing unemployment), or whether the series follows seasons (i.e., smaller unemployment in July). Also, what is very important, the analysis allows for prediction with some confidence of the future state of the phenomenon, so called forecasting. This is because the data are often highly dependent (i.e., unemployment this month depends on the level of unemployment last month). Hence, correlation of the random variables is one of the main features of the time series analysis. This course introduces some of the descriptive methods and theoretical techniques that are used to analyze time series. Many examples will be shown from various applications, such as business and economics, climatology, chemistry, biology. Real data sets will be analyzed using the statistical computing package MINITAB. Computing Time series data are often very large and any sensible analysis requires fast calculations. There are various statistical packages which have many time series analysis options built in. Alternatively one has to write special computer programs. In this course we will use MINITAB and its quite wide set of functions available for time series description and analysis. You need to have an account on the PC network. While logging in remember to choose Time Series from the list of courses. This will allow you to get the data for analysis. Knowledge of MINITAB is not essential (of course it would be helpful). There will be an Introduction to MINITAB session in a computer lab at the beginning of the course. Prerequisites An essential prerequisite is Probability II. Good understanding of bivariate random variables is important to follow the course. Although the course is very much self-contained Calculus II and Fundamentals of Statistics II are advised to be passed. Differentiation skills will be needed as well as understanding the notion of estimation. Students not having passed these two courses may take the Time Series course but should be aware of a necessity to read up on these topics. Assessment 20% in-term, 80% exam paper. There will be six computer labs (every other week, starting in week 2). Attendance to the labs is compulsory. The coursework will be discussed either during the labs or during lectures and solutions will be put on the course website. However, the coursework will not be marked. Instead, there will be two (multiple choice) tests: one in week 7 and one in week 12 of the term. The tests will be based on the material very similar to the coursework. Each test counts 10% of the final mark. Updated on 18 September 2008