*****************************************
Note re laptops:
Participants are required to bring a laptop to this course with their choice of either SPSS, SAS or Stata installed. In most cases university staff and students have access to at least one of these packages via their university. If so, it is important to double check that it works while you are not on campus as sometimes the settings may require access to your university's VPN, particularly for SPSS. If the package works while you are at home, then there is nothing to worry about. If not, you may need to contact the relevant IT department at your institution.
If you do not have access to any of these packages, temporary trial versions of SPSS and SAS are available online (care should be taken not to download these too early). ACSPRI staff will be in touch with all course participants a few weeks prior to the course to make sure there are no issues and/or to help offer solutions.
A wireless network will be available during the course.
*****************************************
The course is designed for those with previous knowledge and experience with regression analysis, who seek to enhance their current knowledge and extend their proficiency in applying the technique. In addition, the course provides the necessary background in statistical methods for those seeking to advance to courses dealing with structural equation modelling (SEM).
The main focus of the course is on the practical side of using regression to address common methodological, statistical and data issues that arise in the social sciences. The course begins with a brief review of the principles of multiple regression. It then examines the specific issues and problems that arise from its application, considers appropriate approaches to address the problems, and how to interpret the results that it produces. These issues include:
• regression diagnostics to detect potential violations of multiple regression's assumptions
• how to identify the presence of multicollinearity and what to do about it
• establishing the existence of non-linear relationships and identifying the transformation procedures necessary to apply to the data
• how to treat missing data, particularly the use of multiple imputation (MI) approaches
• how to incorporate and interpret interactions and moderator variables
• how to model and test for mediating variables, including the classic Baron-Kenny approach, the Sobel test, and more modern approaches
• how to estimate model with binary and ordinal dependent variables (including binary logistic regression and ordered probit)
• how to estimate models with simultaneity (“two-way” causation) using two-stage least squares.
The course concludes by examining the relationship between multiple regression and structural equation modeling (SEM) with latent variables and compares the results of regression and SEM analyses.
Participants will have an opportunity to analyse their own data and discuss the output if applicable.
David John Gow is a consultant in research methods and statistics and their application in the social sciences. He has taught in many ACSPRI Summer and Winter Programs
Knowledge of elementary statistical techniques and some knowledge of the principles of multiple regression at a level comparable to that provided by the ‘Fundamentals of Multiple Regression’ course, and basic knowledge and skills of a statistical package (SPSS, SAS, Stata or R).
The instructor’s course notes, which will be distributed to all participants, serve as the course text. An additional, optional resource is Timothy Keith, Multiple Regression and Beyond (Allyn & Bacon, 2005), which covers multiple regression and basic structural equation modeling.