ARHIV DRUŽBOSLOVNIH PODATKOV

Statistical methods commonly employed by social scientists have become increasingly complex as computers have become more powerful. Often, however, the results of these methods are poorly displayed. For example, a preponderance of published research focuses only on model coefficients even when taken alone—i.e., without further calculations—they have very little intuitive meaning aside from telling us whether a relationship is positive or negative. This practice is even more problematic for models containing interactions among predictors, which results in coefficients that cannot be effectively interpreted independently. The complexity of model coefficients is further magnified when nonlinear models are fitted. In short, model coefficients alone rarely tell the story effectively. With this in mind, this course deals with various methods for effectively displaying results from statistical analyses. In particular, we will demonstrate how relationships, differences, and effects can be clearly communicated using tables and graphs of fitted values derived from statistical models.



The course will start with a discussion about how to effectively display distributions and relationships. Particular emphasis in this regard will be placed on density estimation. We will also discuss general principles for designing good graphs and tables. We will then focus on how to effectively display results from generalized linear models and related methods. Specific topics to be discussed will be substantive significance, predicted values and confidence bounds, contrasts, marginal and partial effects, model fit and selection, and interactions. Similar procedures for multi-level models, especially with respect to understanding effects from different levels, will also be explored. All of the analyses for the course will be done using R, which has exceptional graphical capabilities. No prior knowledge of R is assumed, however.



Costs

* Participants from academic institutions: £610

* Participants from non-academic institutions: £980