FAQ

You describe your need through the contact form, or you book a 30-minute call. Together we scope the question, the data available and the deadline. This step is free and commits you to nothing. I then write you a proposal; nothing starts before you agree.

Not at all, and it is true of most requests. Describe your question in your own words, with the context of your work. Translating that into a statistical question is precisely part of the job — and it is often the most decisive step of the whole assignment.

Yes. I work with health data routinely. Your data never leaves my workstation, is used for no other purpose, and is deleted at your request when the assignment ends. A non-disclosure agreement can be signed before anything is sent. For patient data, send it already pseudonymised where possible.

You do, entirely and exclusively. All code, programs, documents and results developed during our collaboration are transferred to you. You are free to modify them, publish them or hand them to someone else.

It depends on the scale of the work, the state of the data and the deadline — three things I cannot guess before understanding your situation. So I would rather give you a firm figure after scoping, which is free, than a range that would mean nothing. Describe your need and you will get a precise number.

In French or English. The analysis is done remotely, so location does not come into it: I work for clients in Belgium, where I am based, and abroad.

Only scoping, debriefs and training benefit from being in person — easy in Belgium and neighbouring countries, and replaceable by a video call further afield.

I will tell you at scoping, before you commit. It happens: the need belongs to a different profession, or the data cannot answer the question asked. You would rather know straight away, and I would rather say so than deliver a result I do not believe in.

R is an open-source programming language dedicated to data manipulation, statistical analysis and data visualisation.

Being free and open, it evolves continuously: new functions are published daily by the community, whereas proprietary software only moves at the pace of its vendor’s releases. Its very large community also makes it likely that the answer to your problem already exists somewhere.

For three reasons, in this order of importance:

• reproducibility: anyone can rerun your analyses and get exactly your results — including you, six months later;
• automation: the same analyses rerun on updated data without redoing the work;
• extensibility: you can write your own functions instead of being limited to what the tool anticipated.

A spreadsheet remains excellent for quick exploration. It becomes risky as soon as the same calculation has to be redone, checked or handed over.

Size matters less than how often decisions repeat. As soon as you regularly make comparable decisions — which customers to follow up, which site to watch, which treatment to favour — there is value in identifying trends, relationships and opportunities in your data rather than deciding on intuition. You do not need a lot of data for that; you mostly need the right data.

For development and analysis, yes: it is the tool I know best and the one best suited to work with a heavy statistical component. For methodological advice, though, the software is irrelevant: I can support you whether you work in SPSS, Stata, SAS, Python, Jamovi, JASP, JMP or Excel.

Usually not. On datasets of typical size, a classical statistical model performs just as well, explains itself in one sentence and needs no special infrastructure to maintain. I propose a complex method only when it delivers something simple methods do not — and I will tell you which of the two you are looking at.

For anyone who has access to data and would like to analyse it.

Some background in programming or statistics helps, but most courses are designed for beginners and can be followed with no prior experience. The prerequisites for each module are listed in the training catalogue.

As with any language, the key is to keep practising afterwards: that is what separates having seen it from being able to do it.

Learning R by yourself is entirely possible — online courses, books, forums — but it has limits. A course guarantees the trainer’s expertise (technical and pedagogical), the ability to adapt to participants’ questions live, work with current techniques, and gain the most skills in the least time.

The main gain is not the content, which exists everywhere: it is the time you do not spend stuck on a problem someone can solve in two minutes.

A course has to do two things: give immediately usable skills, and give the means to carry on learning alone afterwards.

For that, I believe a trainer has to be three things at once. Competent in the subject, down to its subtleties. A good teacher: able to transmit, to adapt to the participants’ real level, to step back, and to admit what they do not know. And passionate, because someone passionate never stops learning their subject and makes a course markedly easier to follow.

My PhD and daily use of R give me the first; university teaching to very different audiences gives me the second. For the third, you will have to take my word for it — or come and see.

I deliver every course myself; none is subcontracted.

No, it takes place in person. Remote is easier to organise, but interaction collapses: participants ask fewer questions and it is harder to spot who is falling behind. Since that is exactly what makes a course worth more than an online tutorial, I prefer to stick to in-person delivery where it is possible.

You are welcome to propose the venue — your offices work well, provided the necessary equipment is available.

The slides, code, exercises and documents developed during the course are sent to you, so you can go back to them as often as you like. The materials are written in English, but the course itself is delivered in French or English depending on the audience.