In our series Interview With A Team Member, we aim to introduce the people behind the JASP project. Today we are interviewing Koen Derks.
Koen Derks is a PhD candidate at Nyenrode Business University and at the Psychological Methods group at the University of Amsterdam. He is one of the main contributors of the machine learning module, and the audit module
What is your professional background?
I completed my research master in 2018 at the University of Amsterdam with a specialization in Psychological Methods. Due to a growing interest in Bayesian methods and two Artificial Intelligence courses that I did as a minor, I decided to reach out to Eric-Jan Wagenmakers and JASP for internship opportunities. These minor courses fascinated me. One of them was about using nature-based algorithms to solve complex problems (think of ant colony optimization algorithms to find the shortest path in a network), and the other about the implementation of neural networks in real-world scenarios (think creating a self-driving car for a student tournament around a simulated race track). As it turned out, there was a similar project that I could take part in at JASP; the Machine Learning module, which has been released recently. For my Ph.D. I have made a switch from Psychological Methods to the field of Financial Auditing. Currently, I work at Nyenrode Business University, where I research the application of Bayesian methods in the field of auditing. It is interesting and exciting to combine the rules and regulations in this field with Bayesian inference.
What is your favorite statistical test?
My favorite statistical test must be the binomial test, simply because it is so easy to understand and explain to students. I teach a lot of statistics to bachelor students and —as most statistics teachers must recognize— it is not always easy to come up with concrete and understandable scenarios. Coin tossing is a situation that everybody understands, and it lends itself exceptionally well to in-class demonstrations. In a Bayesian setting the conjugate beta prior updates very naturally, making it easy to incorporate Bayesian learning into my lessons.
What do you do for JASP?
Currently, I focus mostly on the audit module that I have created for my Ph.D. project. This module provides a streamlined workflow that auditors can use for audit population testing. I was really excited to make this (frequentist and Bayesian) workflow, as it was not made easily available before. Another project that I did during my studies was creating a module for Bayesian Informative Hypothesis Testing (bain), which is available in the main version of JASP. Lastly, a project that I am especially fond of is the Machine Learning module, which involved many students and programmers, who I cannot thank enough for their help and effort. I hope that the module will be used extensively to promote (knowledge of) machine learning within and outside of psychology.
What feature of JASP do you like best?
This might be a general feature, but I really like the figures and plots that JASP produces. I believe that readable (and beautiful) figures are essential to understand statistical evidence and findings. JASP creates beautiful figures out of the box and offers a bunch of options for exporting them, which is ideal when you are making reports or assignments for students. I even heard that plot editing is in the making! A second feature that helps me tremendously in teaching is the fact that the input and output fit on one screen. Consequently, this removes the need to switch between screens, especially when changing the input options. Furthermore, it’s very nice to see the output respond immediately to input changes.
What aspect of JASP would you like to see improved in a future version?
That must be more extensive data editing facilities. There is just something about editing from a separate .csv file that feels troublesome and non-optimal. For teaching purposes, this would also be a very useful feature as you can enter data on the fly without switching applications.
Are you a Bayesian, a frequentist, an agnostic, a pragmatist, or perhaps something else?
Let’s just say a pragmatic Bayesian. In the audit process, there is so much information that is already gathered during earlier stages. It would just be a waste to not use that information to create an informed analysis. Bayesian statistics offers the unique possibility to incorporate all kinds of audit evidence into the auditor’s design, so I can only argue that to be my preferred methodology.
What question would you like to answer?
In which field would you like to see JASP grow?
I suspect that JASP will fit exceptionally well into the business community. With my current position at a business school, this suspicion is confirmed more and more. Organizations are becoming more and more data-driven and the need for software that makes statistics easily comprehensible is growing. With the addition of the Machine Learning module, I believe that any business can gain from JASP, as it perfectly fits in current trends of Big Data techniques that are now extremely popular.