Author Spotlight
In the Author Spotlight series, TDS Editors chat with members of our community about their career path in Data Science, their writing, and their sources of inspiration. Today, we’re thrilled to share our conversation with Hannah Roos.
Hannah Roos is a psychologist with a passion for data science. Inspired by her work in management consulting and behavioral science, she writes hands-on coding tutorials and deep dives in R and Python. She is driven by the mission to actively bridge the gap between evidence-based research and the business world.
What was your entry point into data science?
During my bachelor’s in Psychology, we were taught about the importance of a sound methodological approach to research questions, especially in the "softer" sciences, which got me interested in the part most students fear most: statistics and hypothesis testing.
I realized that our minds are probably not wired to understand these things intuitively, but this is what actually sparked my ambition: I wanted to understand how you could solve questions by means of data.
How do you approach the more difficult topics within the field?
I love the challenging aspects of data science because once you have mastered a task after many hours of deep thought and hard work, you get a great sense of accomplishment and you can finally see the beauty behind all the math.
During the end of my studies, I worked as a research assistant for experiments on visual cognition. All my co-workers seemed to be more skilled at programming and complex data analyses: this was probably a particular challenging part of my journey. It almost seemed like they had access to a parallel world full of logical ideas, and they had the right vocabulary to debate them. But this world somehow remained closed to me at that stage. Maybe I was simply not born for this in the first place?
But instead of focusing on whether I was "smart enough" or not, I have kept on learning to become more and more skilled. By now, I have completed more than 60 courses on Datacamp and extended my knowledge about R, Python, and SQL, as well as common concepts in statistics and machine learning.
Reading your articles, one gets the sense that your psychology background often informs your work.
Yeah, that’s true! As a psychologist, I always try to find out what makes people tick by means of data, and this is not a straightforward task.
I have a sense for actual problems that occur within the interface of psychology and the business world, and these really inspire my writings. For example, I have recently written an article about using AI to predict employee turnover in companies. Of course, you could build a machine learning algorithm to classify employees into 1) those at-risk of leaving and 0) those who are likely to stay. But honestly, there are so many aspects that influence one’s motivation to work at a specific company, which makes it difficult to generalize beyond in-sample predictions – for example, the relationship with your supervisor, market dynamics, learning opportunities… all of which can change within a relatively short period of time. So how could you accurately predict such a complex decision in practice?
While I am convinced about using data to drive decisions in the real world, I think we need to use our common sense and domain knowledge, too. This is how we can confidently say what the data can tell us about the future – and what it can’t.
What are the kinds of problems you’re most motivated to solve through your work with data?
I love data-focused projects that ideally serve my curiosity as a scientist as well as the broader public. In general, I would like to dive a bit deeper into psychometric testing, that is, using statistics to improve ways to measure traits that we cannot measure directly (e.g., personality). This has the potential to tell you something about how people differ on a very profound level.
Currently I’m analyzing data on the psychological impact of COVID across the early days of the outbreak in 2020 – this is why my next article will include aspects of time-series analysis and coding examples in Python. From a broader perspective, I am interested in analyses across time to investigate change and development.
What prompted you to start writing publicly about your work and other data-related topics?
I started writing about the p-value and what it really means to conduct hypothesis testing. That was because I realized that I had gotten it so wrong for a long time, and felt like there was still a lot of misunderstanding around this concept from students and professionals alike. I wanted to share the "AHA!" moment I finally had with others who also struggle: what if I could write an article that would actually make it easier for people?
I like to write about data-related topics because it is amazing to work on something that I am passionate about, but I also want to live in a world where there is a free sharing of knowledge. Hopefully I can contribute to such a world this way.
It looks like you’ve made a conscious choice as a writer to aim for thoroughness and depth over output volume. What does the process behind your articles look like?
I want to provide a complete and holistic view on a problem instead of rushing through the programming part. Usually, I start with a concept that I would like to understand in more depth—let’s say Bayesian statistics or structural equation modelling. Then I start reading some scientific papers and look for a public dataset that I can use to experiment with in parallel.
Once I feel like I have derived some key insights from my learning project, a good use case, and an efficient coding approach that others could benefit from too, I actually start writing. To make it easier, I start with bullet points on a plain word file that contains the critical points I want to make, and then replace it with a coherent text step by step.
Do you have any practical advice for data scientists who are toying with the idea of writing about their work?
For beginners, I would say that you do not need to prove anything with your article. Maybe just explore a topic for fun and write down your insights. Imagine that you would tell these to your friend who does not actually care much about data science but still listens to you politely. Ask yourself: Can I spark a bit of interest anyway? If so, you probably have a good story to tell. If not, you at least had your very own little voyage of discovery.
Data science is a field in constant flux—are there any particular changes you hope to see in the coming months and years?
I would like to see data science being even more inclusive and accessible to everyone – instead of asking experts from industry and academia, I think that everyone should have the chance to become such an expert themselves and derive their own insights from data.
Hopefully, a term like "AI" won’t elicit the idea of an ominous black box that exploits our personal information to make ever more profit – in an ideal future, people from the broader public would finally understand how algorithms work to assess for themselves to what extent they could be beneficial.
For the broader tech world, I would like to see more automation for tedious workflows so people can focus more on the things that actually matter to them. I believe that some tech startups already have great data-related innovations that are able to revolutionize the way we live, and allow us to become even more autonomous individuals – when designed and used in a thoughtful way.
You can follow Hannah on Medium to stay up-to-date with her latest work; for a taste of Hannah’s past TDS deep dives and tutorials, here are a few highlights:
- Uncover the Latent Structure of Personality with lavaan in R (January 2022, 29 minutes)
- Smart Handling of Missing Data in R (February 2021, 15 minutes)
- Data Storytelling for Survey Data in R (December 2020, 8 minutes)
- Better Done Bayesian? (April 2021, 30 minutes)
Feeling inspired to share some of your own writing with a wide audience? We’d love to hear from you.
This Q&A was lightly edited for length and clarity.
