WA102 - Introduction to Conjoint Experiments

Estimating causal effects is a central aim of quantitative empirical analysis in social sciences. Recently, Conjoint Analysis and Choice-Based Conjoint Experiments have gained interest among social scientists to understand and predict people's preferences in a multi-dimensional and multi-choice environment. This course offers an applied introduction to Choice-Based Conjoint, along with hands-on experience in lab sessions. 

By the end of the course, you will:

1. Have a basic understanding of the structure, logical underpinnings, basic notions, and analytical goals of conjoint analysis. 
2. Identify areas of application where conjoint analysis could be successfully implemented. 
3. Critically evaluate conjoint experiment applications and understand the advantages/disadvantages compared to traditional methods. 
6. Implement your own conjoint experiment into an (online) survey platform.
7. Understand and be able to apply different techniques to analyse conjoint experiments. 
8. Be able to efficiently visualise your results 
10. Be well prepared for more advanced conjoint and factorial design courses or workshops. 

The course is structured around six key topics:

1. I present the general idea of conjoint experiments. I introduce the logic underlining conjoint experiments, their development, and the reasons behind the rising interest in them within the social sciences. 
2. I briefly introduce the 'utility' function at the base of Choice-Based Conjoint analysis. In particular, I give an overview of the Random Utility Theory and discuss it within the framework of social and behavioural science. 
3. I present different ways to measure individual preferences in a conjoint experiment. The focus will be on Choice-Based Conjoint measurement but we will briefly discuss other measurements (e.g. Rating, Ranking, Combined, and Adaptive). 
4. I give an overview of different types of conjoint design, their use, and their limitations. I present and explain Alternatives, Choice Sets and Context design, paying particular attention to the design of conjoint alternatives. 
5. I focus on the construction and implementation of a conjoint experiment. Specifically, I show how to design alternatives with attributes/levels constraints and randomisation using a ready to use JavaScript/Python program and R. 
6. I give an overview of different methods to analyse a conjoint experiment. Specifically, I focus on AMCEs, marginal mean and omnibus F-test. We will briefly discuss subgroup differences and visualisation. 

Note The goal of this workshop is to give an applied introduction to conjoint experiments. If you are already familiar with conjoint analysis or you are interested in the broader theory behind conjoint and factorial experiments, this is not the right course for you.

Please bring your laptop with the latest version of R (3.6.x, Planting of a Tree), R Studio (1.2.x) and Python 2.7 installed. Note that Python 2.7 and Python 3.7 are not the same. Unfortunately, the softwere required by the randomisation mechanism is still running on Python 2.7.

Make sure they work and that you can run a script before coming to class since we will have no time to resolve technical issues. If you have already collected data, bring it along. If not, you’ll get a toy dataset to play with. Be sure to have installed in R the cjoint and cregg packages together with any other package that you use for data management/cleaning/visualisation (e.g. dplyr,ggplot, etc).

This course will use R, which is a free and open-source programming language primarily used for statistics and data analysis. Although you are allowed to use other solutions, we will also use RStudio, which is an easy-to-use interface to R. 

If you encounter any problems with R Studio on your local machine, use R Studio Cloud.

Please bring your laptop.

There are no specific requirements other than being able to use a browser (for dataset downloads and R Studio Cloud) and having installed R Studio and Python 2.7.