We are announcing three short courses, which are scheduled to take place in February and March 2007 at The University of Reading, UK. Summary information is given below. For more detailed information and registration forms please see http://www.reading.ac.uk/ssc/ providing your address and/or fax number, or email [log in to unmask] Multilevel Modelling ********************** Dates: 26-27 February 2007. Duration: 2 days. Price: 585 GBP. Hierarchical data structures are common in many areas of application, including the social sciences, education, market research, biology, psychology, agriculture and industry. Such data structures may arise from surveys, designed experiments or observational studies. In the past, because suitable modelling approaches were not available, the analysis of hierarchical data was rather ad hoc and, in some well-known cases, ignoring the multilevel nature of the data has even led to incorrect conclusions. Although the theory of multilevel modelling will be explained, the emphasis in this course is on its practical implementation and interpretation of results. The course will focus on models with normally distributed errors. The MLwiN package will be used in the presentations as well as other software as appropriate. In the practicals, in addition to MLwiN, participants will be able to choose from the statistics packages R, SAS and Stata. Multilevel Generalised Linear Models ************************************** Date: 28 February 2007. Duration: 1 day. Price: 315 GBP. This course extends the traditional multilevel model with normally distributed errors to those with discrete responses. Common discrete variables are binary and counts. Typically these types of responses are analysed using generalised linear models such as logistic regression and Poisson regression. Commonly used generalised linear models will be extended to deal with multiple error structures, using a variety of examples. The emphasis will be practical, although an outline of the theory will be presented. Examples will be analysed using the specialised multilevel package MLwiN and other software such as SAS PROC NLMIXED. In the practicals there will be a choice of R and Stata as well as SAS to complement the use of MLwiN. Generalised Estimating Equations - What, Why and How ****************************************************** Date: 1-2 March 2007. Duration: 2 days. Price: 585 GBP. The methodology of generalised estimating equations (GEEs) was developed by Liang and Zeger for analysing discrete longitudinal data. This course will introduce GEE methodology and how it fits in with other modelling techniques. The use of GEEs for modelling non-normal correlated data, such as repeated measurements, will be covered and emphasis will be placed on how the methodology can be implemented. Fitting GEEs will be demonstrated during lecture sessions, and participants will have the opportunity to fit models themselves. Examples using count data and binary and categorical data will be given. The GENMOD procedure in the statistical package SAS will be used on the course. Julia Harris Short Course Administrator Statistical Service Centre Tel: 0118 378 8689 Fax: + 44 (0)118 975 3169 email: [log in to unmask] Julia Harris Short Course Administrator Statistical Service Centre Tel: 0118 378 8689