Data Science Radar – Modeller Profile

by Henrik B Nyberg, Mango Solutions

1. Tell us a bit about your background in Data Science.

During the final year of my master in chemical engineering I steered away from lab work and the large scale industrial applications. Computers seemed much more compliant and easier to work with than cells so I took some courses in pharmacometrics and I found myself really enjoying exploring clinical trial data. An opportunity appeared for me to work as a consultant at Mango while simultaneously doing a PhD in pharmacometrics and I jumped at it. Working at Mango has then obviously widened my horizons of data science significantly.

2. How would you describe what a Modeller is in your own words?

To me, a modeller is someone who doesn’t necessarily care too much about the data, but really wants to get to the bottom of the mechanisms behind the data.

3. Were you surprised at your Data Science Radar profile result?   Please explain. 

I am already profiled towards modelling so my characterization as a modeller was more of a relief than a surprise. It is really interesting though to look at the results we had in the team. I can definitely trace many characteristics of these roles to different strengths and different viewpoints that the different consultants take when looking at a new problem.

4. Is knowing this information beneficial to shaping your career development plan?  If so, how?

Absolutely! I think that the fields where I scored low are especially beneficial. I think I will try to spend some more time with my communicator and technologist colleagues to see what I can learn from them.

5. How do you apply your skills as a Modeller at Mango Solutions?

In part I apply these skills directly for building models that describe data. Another part of my work is  much more technical in nature. It involves understanding the workflows and needs of people like me and translating that into feature requests for the different solutions that we build for customers.

6. If someone wanted to develop their Modeller skills further, what would you recommend? 

As I am not a statistician I initially felt unsure of some of the theory behind things like likelihood and model fitting. I have made an effort to really dig into some of this theory and it has proven tremendously beneficial. If you are just starting out I would look at some example applications of the lme4 R package. R is quite a powerful modelling tool as long as your data is a sensible size, and you don’t need to solve differential equations or such.

7. Which of your other highest scoring skills on the Radar compliments a Modeller skillset and why?

Visualization, my second highest score, is my most important tool for communication. The conclusions and descriptions of mechanisms that I deliver are often quite complex and something that I have spent a lot of time analyzing. To quickly convey the major points of my analysis is so much easier when I have good quality graphics.

After modeller and visualizer I am a programmer. The programming I do is mostly using R to build automated workflows and is something that I feel really makes my work more efficient. I often use R to build workflows where multiple versions of models are run in specialist tools such as Nonmem and Monolix, and the results are evaluated and used for the next step in the workflow.

Do your own Data Science Radar here