Methods and statistics 1: Regression analysis and its generalizations

Students will:

1. Understand and interpret generalized linear models, including a thorough understanding of the assumptions that need to be met for their application as well as the consequences of minor and major violations of such assumptions.
2. Apply generalized linear models to unfamiliar social science research questions as well as perform hypothesis tests and model selection.
3. Acquire practical skills in data handling with statistical software R using syntax.

In the first part of the course, you will work through a syllabus with an extensive set of assignments on working in syntax and data handling (description of data, construction of variables, use of functions, aggregating data, restructuring data, combining datasets, etc.). Factor analysis and reliability will also be taught.

In the second part of the course, we start with a two-week discussion of univariate (one dependent variable) linear models for continuous response variables with one or more explanatory (independent) variables that may be continuous or categorical (dummies, ordinal, unordered/nominal). We pay some attention to the requirements for causal interpretations of regression, but do not discuss formal methods for causal inference. We extensively discuss model building and moderation (interaction) analysis. We then switch to regression-style models for types of dependent variables for which the ordinary regression model is not suitable. These include binary variables (e.g., membership of a union), nominal variables (e.g., denomination of school attended), and ordinal discrete variables (e.g., an item measuring intention to participate in an election using 4 ordered answering categories). Finally, we discuss a popular approach in event history analysis, explaining how the time-till-the-occurrence of an event (e.g., the birth of first child, the divorce of a marriage, promotion in a career) can be modeled in terms of predictors.