“Astrostatistics is a field full of opportunities right now”: An interview with Roberto Trotta

Dr Roberto Trotta is a theoretical cosmologist in the Astrophysics Group of Imperial College London, where he is a Reader in Astrophysics (equivalent to Associate Professor in the US), and the Director of Imperial’s Centre for Languages, Culture and Communication.

As Director of the Centre for Languages, Culture and Communication, he is part of Imperial’s senior management team and provides academic lead and strategic direction for the Centre’s activities: the Imperial Horizons programme (delivering teaching in humanities subjects and languages to over 4500 undergraduates at Imperial); the Evening Classes programme; and two MSc programmes in Science Communication.

Dr Trotta is also a science communicator and takes part in numerous public engagement with science activities, from science festivals to radio broadcasts. His first book for the public, The Edge of the Sky: All You Need to Know About the All-There-Is(Basic Books) explains the Universe (‘All-There-Is’) using only the most common 1,000 words in English. For his book, he  was named by Foreign Policy one of the 100 Global Thinkers 2014.

Dr Trotta gave this year’s Significance lecture at the Royal Statistical Society’s annual conference in Glasgow, where Alison Oliver talked with him about his career.

1. You obtained an MSc in theoretical physics from ETH Zurich and then a PhD in physics from Geneva. What was it that introduced you to the discipline in the first place and inspired you to pursue it as a career?

I studied physics because I wanted to learn how the universe works, and this is also why I did a PhD in theoretical cosmology. Early on I became interested in the data analytics aspect as then — in the early 2000s — the great big data sets in cosmology were really starting to come together fast. There was a real need for new methods, activities and methodologies to understand, analyze, and interpret these data. That caught my interest and my imagination: rather than work in theoretical physics with models that nobody could possibly test or disprove, you could now bring those theories together with actual data and get an answer—probabilistically of course. So you’d get something out that would either show that the data and models actually were compatible or else kill models that were not supported by the data. That is what got me to study statistics: being able to test sometimes very outlandish ideas about the fundamental nature of our universe, and come up with an actual answer.

2. You then moved to Oxford University as Lockyer Fellow of the Royal Astronomical Society. What inspired this move?

It was a wonderful opportunity after my PhD. I had applied for this very prestigious fellowship, was lucky enough to get it, and took it to Oxford, which was a fantastic place to be, because of all the interactions with all sorts of academics. Plus, the collegiate life, and the astrophysics department were excellent.

The beauty of Oxford as a place is that there’s space and opportunity for cross-disciplinary conversations. You are not locked into your own physics world; you can have conversations with philosophers, with statisticians, with people studying English literature, which is really inspiring – I’m interested in science communication, education, and statistical methods. I enjoy being able to talk about very different intellectually stimulating subjects and that was in abundance at Oxford.

3. Amongst your many roles, you also are co-founder and director at Data Fusion Consultants. What does Data Fusion offer?

Data Fusion Consultants is a spin-off company that I founded to apply the methods that I’ve developed for my research in statistics to problems in the real world – so big businesses looking for data analytical consultancies, or governmental offices that need solutions that they cannot provide in-house. It’s a way of taking research that has been developed to solve problems in fairly –shall we say– esoteric domains, such as cosmology, and apply it to many problems in the real world that are quite similar from a methodological and structural perspective. So you can take these very advanced, fancy methods that are used to investigate say dark energy and apply them successfully to difficult but fascinating and important problems that can benefit the entirety of society.

4. Many leading experts in statistics have insisted on the importance of collaboration with other fields and astrostatistics would appear to be an excellent example and quite a new discipline in its own right. How are you personally finding how statisticians and astronomers work together?

After a while…very well! In the beginning, it is difficult, mostly because of language barriers. The difficulty comes from the fact that you have to blend astrophysical insight, knowledge, models, etc. with statistical techniques, and statisticians have to learn the astrophysicists’ jargon, and vice versa. Astronomers sometimes reinvent the wheel to solve problems that statisticians have known how to do for decades, while astronomers are still working on such issues. But that’s precisely why the collaboration can be so fruitful: because things that are widely established in statistics can be brought to bear on problems in astronomy that astronomers don’t really know how to solve yet.

Vice versa, sometimes we’ve seen in the past couple of decades that problems in astrophysics and cosmology have really motivated a spur of development in statistics – mostly applied statistics, computational statistics, and computational methods. Such methods have been developed by astronomers to solve their problems, and then were found to apply elsewhere. It’s a really good collaboration if you can get it going. In order to do that, you need a time investment and two teams coming together—statistics, astronomy—to work together for a number of years. We have a great collaboration now well established at Imperial with Professor van Dyk, who is Head of Maths there. This has taken us two or three years to really get going. Now we have joint students and post docs, we work together, publish together – it’s working very well. But it takes a big investment in the beginning from both sides to take a problem outside of its astrophysics context and trim it down to a level that can be discussed in its essential form – trimming away all the jargon, all the stuff that’s only astrophysics, that statisticians don’t really care about – and then transform it into something sufficient to work with, and then you’re really in a position to get going. With very good results!

5. Your research in cosmology focuses on a goal to learn more about the history and nature of the Universe. What are you working on currently?

Right now, I’m working on statistical and theoretical aspects of two of the biggest questions in my field: one is the nature of dark energy, which makes up about 70% of the universe, the other is the nature of dark matter, which makes up another 25%. So these two things make up 95% of the universe, and we have no idea what they are! To make advances in both of these domains, we need data-driven methods to characterize both dark matter and dark energy in the face of a multitude of data sets that we are now gathering from a variety of telescopes and other detectors. On the dark energy side, I am working on new methodologies to better analyze present, and upcoming data of stellar explosions known as supernova type Ia that help us measure how fast the universe is expanding and how fast it has been expanding in the past. Dark energy supposedly has a negative pressure, that is ripping the universe apart, effectively. In order to understand what it is, we need to measure the expansion of the Universe more precisely, and to do that we need to better understand the data and the models themselves. Dark matter is thought to be a new type of particle, which we can hopefully observe in a lab, as well as in the sky, albeit indirectly. My work is about combining these three very different data strands: cosmological data, astrophysical data, and data from particle physics experiments into a unified whole that can tell us more about the nature of dark matter.

6. Your lecture at RSS 2017 was the Significance Lecture. What was the one thing that you would like your audience to take away from your lecture?

The most important thing is that astrostatistics is a field full of opportunities right now. We’ve got this big data coming—they’re not just big, they’re also very complex and have an interesting structure. Already now, our understanding of the universe is being limited not by the amount of data, but by our ability to interpret, manipulate and make inferences from those data. So really the stumbling block that’s limiting our quest for knowledge about the fundamental nature of the universe is the analysis of data, and will be even more so in the future. So it’s a field that’s ready for new ideas and for more people to work in it. For statisticians looking for complex, challenging and interesting problems, astrostatistics is the place to go.

Astronomers sometimes reinvent the wheel to solve problems that statisticians have known how to do for decades, while astronomers are still working on such issues. But that’s precisely why the collaboration can be so fruitful: because things that are widely established in statistics can be brought to bear on problems in astronomy that astronomers don’t really know how to solve yet.

7. What do you think have been the most important recent developments in astrostatistics and will these influence your teaching in future years?

Probably the computational methods that we’ve developed to deal with these data sets. These computational methods are important because firstly they solve the problems that we are after, the scientific problems of the day—I mentioned dark energy, dark matter; we haven’t solved them yet, but we’re working on it—but also because they can be translated into more general algorithms and ideas that can be used for more general problems outside of astronomy.

So perhaps the biggest of my “problems”, if you like, as a statistical consultant at Imperial, is that people don’t realize that they can go to an astronomy department asking for help for a data analytical problem, because of course they wouldn’t necessarily think of data analytics being big in astronomy, but it is. The fact that this fundamental resurgence, largely funded by governments and grants, can actually have very important spin-off applications in very applied fields, is something that people often don’t realize. I believe that these applications of very fundamental research to real-world problems are one of the most important developments that we’ve seen in the last 20 years.

8. Your research has been published in many journals and books: is there a particular article or book that you are most proud of so far in your career?

There is a paper I wrote in 2007 called “Bayes in the Sky” which was a review article that introduced some of the concepts that I’ve learned about particular branches of statistics—Bayesian statistics from the statistics literature—and then applied them to cosmological problems of the day. That turned out to be quite influential, because many people realized those tools and ideas could actually be used and exploited for our problems. It was a little bit ahead of its time and now it’s become mainstream, so it’s good to see these ideas and methods now being applied to all sorts of problems in the field.

9. You are the recipient of numerous awards including the Lord Kelvin Award of the British Association for the Advancement of Science. Is there an award that you are particularly proud of?

In 2014, Foreign Policy Magazine named me as one of the 100 global thinkers for the year which was a great honour. Receiving recognition for my book, The Edge of the Sky was also an honour, as I had written it for the general public, in 2014 explaining cosmology and the big mysteries that we’re working on today. In the book, I used only the most common thousand words in English, of which “universe” is not one, so it was a little bit of a radical science communication experiment that was quite successful. I think the success of the book is due to the fact that it was quirky and different and it spoke to an audience who would not necessarily come and listen to an astrophysics or cosmology lecture, because they might be intimidated by the subject and think it’s not for them. But if you cast it in a simpler way, and a narrative and a story, which the book tried to do in this quirk format with the 1000 words, I think it works.

The book aimed to talk about important questions in science today and reach an audience who wouldn’t necessarily otherwise be exposed to them. I think I’ve been proud of the book and recognition from Foreign Policy because it proved it was possible for a wider cross-sections of society to be brought into the fold of this interesting research and questions about the fundamental origins of the universe—what is the cosmos made of, where do we all come from—those are too important questions to be left to the scientists alone. Everybody’s got a stake in these questions. It’s good to see that the public at large can participate in this kind of research that is funded by public money. It’s only fair that everybody should share in the knowledge that’s being generated.

10. What is the best book in astrostatistics that you have ever read?

I guess the best book in astrostatistics is probably the one I’m writing at the moment, but it’s not out yet! Joking aside, there are a number of books in astrostatistics that are good, but all of them explore particular areas. In fact, the reason for wanting to write a book on the subject is because I don’t think there is a book yet out there that does all the things I would like to see—a textbook for graduate students, for advanced users, and so on. So while there are good books, none of them really has the selection of topics that I think are most important for practitioners today.

In terms of the book that has influenced me the most, it’s not particularly an astrostatistics book, but is the book by the late David MacKay, “Information Theory, Inference and Learning Algorithms” from 2003. This was for many people, not just for myself, an extremely influential book, because it presented ideas about probabilistic inference, model selection and graphical models. It didn’t have a particular focus in astrostatistics, but there are many researchers in my field that ended up reading the book and were inspired and influenced by it and then used those ideas in their own work. So I think to me that was a very inspirational book.

11. What would you recommend to young people who want to start a career in astrostatistics?

Astrostatistics provides a great opportunity to work on problems that are interesting, deep and solvable. These are not unsolvable problems; we can and will make advances thanks to upcoming new data sets, and develop methods that will solve current problems, thus giving young people and their ideas recognition and the opportunity of truly advancing the knowledge in the field. If academia ends up not being the right path for students and young researchers, and they end up leaving astrostatistics as an academic endeavour, with the tools and knowledge that they’ve developed in terms of data analytics, statistics, machine learning, and so on, they become very highly employable. We have a large number of students from our group at Imperial setting off in the most interesting data analytical directions after they’ve been with us. I think it’s a very good direction, and it’s one that has got an assured future both in academia and for society in general.

12. Who are the people who have been influential in your career?

In terms of thinkers, certainly my PhD supervisor, Ruth Durrer, at the University of Geneva, is somebody who very early on in my career really modeled for me what it is to be a research scientist. So in that respect, she was an early influence on my career in that she taught me how to do research. When you’re a PhD student, you don’t really know how to do it, and so it’s very much akin to a pupil learning a craft from a master. To be a scientist, sometimes you can’t learn from experience, as what you are doing is by definition new, so you have to learn by example, and Ruth was my leading light.


Copyright: Image appears courtesy of Dr Trotta