BayesCamp


Small-group training and one-to-one coaching for data visualisation and Bayesian analysis.


Training
We provide bespoke on-site courses in Bayesian models or data visualisation, from beginners to specialists. A half-day session can demystify and explain the jargon for managers, recruiters and researchers. A one-day session will give a broad introduction for beginners and get them started in one of the main software packages. Browse through the specialised topics below, or the future public courses, or email robert@bayescamp.com to discuss your team's goals.


Coaching
There are training courses, websites and books aplenty to learn data science skills, but if you are serious about building a career, you need some one-to-one coaching. Your coach is not a tutor, but someone experienced in data analysis, trained in coaching techniques and aware of the contemporary data workplace. When they ask you the right questions, you can start to discover your priorities, strengths and weaknesses. All coaching is completely confidential. Build a plan to get yourself ahead of the pack, and monitor your progress. email robert@bayescamp.com to discuss your personal journey.


BayesCamp
Throw out everything you thought you knew about data analysis, and start thinking about it in a new, focussed, successful way. Understand and solve each problem as it comes, rather than picking old tools out of the box. Build effective models and know when and why they will work. Stand out as a confident, thoughtful data analyst. Join us for an immersive three-day BayesCamp in the country, away from distractions, with your fellow future leaders in data science. email robert@bayescamp.com to find out more.


Analysis


Structural equation models
SEMs and related latent variable models put your data in the context of the theory and understanding that you already have. Unobservable latent variables drive the observed manifest varibles in ways that are naturally understood, explained and communicated. Bayesian methods are a perfect fit to these methods because estimating the latent variables is done with probability like any other parameters. Combine that with estimation of missing or imprecise data, and hierarchical structure, seasonality or geographical details, and you have a world-beating methodology. Learn how to think it through, implement it in Stan through R or Python, and explain it to others.


Meta-analysis
When you combine statistical results from multiple studies, there's a lot of things that can and do go wrong. You can include all of those problems in a Bayesian model: reporting bias, publication bias, measurement error, dichotomania, different measures, different study designs, indirect comparisons. Get reliable and realistic inferences and predictions. Or you can brush them under the carpet and pretend it's all fine. Your choice.


Non-parametric models
Bayesian non-parametrics are not just a trendy, must-have skill, but also offer a combination of the non-linearity of black boxes like deep learning, without the mystery. Get flexible, data-driven predictions by combining many models in one Bayesian framework. Let the data shine through without imposing some unrealistic assumptions. Learn how to do it, how to critique it, and how to communicate it.


Adaptive clinical trials
Get the skills you need to cope with the increasing demand for adaptive designs in new clinical trials. Learn how Bayesian methods allow flexible analyses with statistical results that are easy to communicate and meaningful to clinical audiences and policy makers.



Software


Stan
The world-leading Bayesian software comes with interfaces for R, Python, Julia and more. It is tried and trusted for almost every kind of model you might want to fit to your data. It is used by every serious data science organisation you can name. But maybe you find it a little scary, a little unapproachable. Let's break through that barrier with an introductory or specialist course.


R
There are many ways to access powerful Bayesian tools in R, from user-friendly, high-level packages like brms and rethinking through to the NUTS and bolts of Stan, JAGS and BUGS. An overview course will show you the options and let you weigh up which one works best for you and your data.


BUGS / JAGS
Bayesian inference Using the Gibbs Sampler is still the world's favourite Bayesian software, and Just Another Gibbs Sampler is close behind. An introductory course will get you up and running with these tools and show you many of the models that have been effectively programmed in them already.



Company principles

  • We focus on a few areas of specialist strength: Bayesian analysis and data visualisation. We don't trick clients with hype.
  • We contribute to local tech and science communities, and prioritise face-to-face, local work.
  • Our work must promote healthy, happy, sustainable workplaces and professional practices for people working with data, whether they call it statistics, data science, machine learning or artificial intelligence. We help people to confront and change stressful, unrealistic, uninformed expectations.
  • We will work with any individuals or organisations based anywhere in the world (unless prevented by UK law, and except any foreign government); dialogue builds a better world. No BayesCamp activity ever physically takes place in jurisdictions that endorse torture or extrajudicial killings, or that legislate discrimination. (Note that, at the time of writing, this unfortunately includes the United States of America.)
  • We never use exploitative subcontracting practices for individuals, such as zero-hour contracts. If we're happy to have someone supply BayesCamp training or coaching, we support them with employment and the legal protection that brings them.
  • Prizes and awards are inimical to nurturing new ideas and people early in their careers. BayesCamp and its directors never accept any prize.


BayesCamp Ltd is incorporated in England and Wales, number 10666858. The term 'BayesCamp' and the Gaussian tent logo are registered trademarks.