This class is envisioned as a class on Intermediate Statistical Data Analysis for Social Sciences. In it we will present concepts and methods that are widespread and important for a successful career as an academic. The class is geared towards non-statisticians and puts an emphasis on “doing” and computation/simulation (as opposed to theory). It can in principle be done with little statistics knowledge, but it helps tremendously if the students had some exposure to statistical thinking and methods in their undergrad studies (e.g., regression analysis).

Throughout the course we will make use of the statistical programming language R. It is highly recommended to be familiar with R to a certain degree. If not, then students are expected to catch up within the first few units.

Here is the course plan (numbers refer to topic units, not necessarily actual units):

[1] We start with an introduction to the R environment and attempts on coding with R.

[2] The next topic is Monte Carlo simulations and resampling, which covers exploration of probability distributions, how to simulate data from these distributions and why that is useful in data analysis.

[3] Next we talk about statistical decision theory and inference, including the concepts of p-values and confidence intervals.

[4] This topic covers resampling/simulation-based approaches to statistical inference.

[5] Next we talk about measuring association between two variables, including the concept of correlation.

[6] In this topic we lay the ground for the principal method of statistics, regression analysis.

[7] In this topic we discuss two-sample t-test and ANOVA with and without repeated measurement.

[8] In these units we cover the general linear model as the unifying framework under which we can subsume everything from Pearson correlation to linear regression and ANOVA.

[9] In this unit we will cover effects sizes, power and the role that the sample size plays for planning.

[10] In this unit we extend the general linear model to the class of mixed effects (aka hierarchical aka multilevel) models.

[11] Next we extend regression analyses to models that follow other distributions. These models are called generalized linear models.

[12] Our last topic unit covers other regression topics including regression with data transformations, polynomial regression and robust regression.

[13] This is the unit where we wrap everything up and connect the dots of what we learned in a big picture view.

The grading is based on the following components (100 points overall):

The students have to hand-in a midterm (50 points) and a final project report (50 points) both in the form of a data analysis report.

In both of these projects, students work in teams of two and must apply the statistical methods we learned in class with the R functions we learned, indicating their mastery of the topics. Ideally, students will use a data set that is of relevance to their own PhD project.