One day of tutorials and two days of talks, open discussions, and statistical modeling in beautiful Helsinki, Finland.

August 29-31, 2018 • Aalto University, Helsinki, Finland

Sponsors
Location
Overall Schedule
Videos
Speakers and Schedule
Posters
Tutorials
StanCon 2018 Helsinki Organizing Committee

Stan (http://mc-stan.org) is a probabilistic programming and statistical modeling language used by tens of thousands of scientists, engineers, and other researchers for statistical modeling, data analysis, and prediction. It is being applied academically and commercially across fields as diverse as ecology, pharmacometrics, physics, political science, finance and econometrics, professional sports, real estate, publishing, recommender systems, and educational testing.

If you have questions regarding registration, payments, travel, please email stancon.helsinki2018@tavicon.fi.

Slides, notebooks and code

StanCon talks repository

Videos

Location

Aalto University, Otaniemi Campus, Maarintie 8, Espoo, Finland

Tutorials on Wednesday 29th August, will be at Aalto University Otaniemi campus (Maarintie 8, Espoo) at 13min metro ride away from the Helsinki city centre.

Aalto University, Töölö Campus, Runeberginkatu 14-16, Helsinki, Finland

StanCon 2018 Helsinki will take place 30-31 August at Aalto University Töölö Campus (Runeberginkatu 14-16, Helsinki) in the proximity of Helsinki city centre.

Finland and Helsinki travel information

Finland and Helsinki Travel Information

Overall Schedule

	Wednesday 29 August, Aalto University, Otaniemi campus, Maarintie 8, Espoo
08:00 - 14:00	Registration
08:30 - 17:30	Tutorials
	Wednesday 29 August, Musiikkitalo, Mannerheimintie 13 A, Helsinki
18:30 - 21:00	Registration
19:00 - 21:30	Get together
	Thursday 30 August, Aalto University, Töölö campus, Runeberginkatu 14-16, Helsinki
08:00 - 17:00	Registration
08:45 - 18:00	Conference
	Thursday 30 August, TBD
19:30 -	Dinner
	Friday 31 August, Aalto University, Töölö campus, Runeberginkatu 14-16, Helsinki
08:00 - 10:00	Registration
09:00 - 17:00	Conference

Speakers and Schedule

Wednesday 29th August

Location: Aalto University Otaniemi campus, Maarintie 8, Espoo

08:00 - 8:30 Registration

08:30 - 10:00 Basics of Bayesian inference and Stan, part 1 Jonah Gabry & Lauren Kennedy. Hall AS2, 2nd floor.

08:30 - 10:00 Stan C++ development: Adding a new function to Stan, part 1 Bob Carpenter, Sean Talts & Mitzi Morris. Hall TU1, 1st floor.

10:00 - 10:30 Coffee Break

10:30 - 12:00 Basics of Bayesian inference and Stan, part 2 Jonah Gabry & Lauren Kennedy. Hall AS2, 2nd floor.

10:30 - 12:00 Stan C++ development: Adding a new function to Stan, part 2 Bob Carpenter, Sean Talts & Mitzi Morris. Hall TU1, 1st floor.

12:00 - 13:00 Lunch Those attending the StanLadies lunch, please meet with your lunch in AS4 (1023-1024)

13:00 - 14:30 Hierarchical models, part 1 Ben Goodrich. Hall AS2, 2nd floor.

13:00 - 14:30 Ordinary differential equation (ODE) models in Stan Daniel Lee. Hall TU1, 1st floor.

14:30 - 15:00 Coffee break

15:00 - 16:30 Hierarchical models, part 2 Ben Goodrich. Hall AS2, 2nd floor.

15:00 - 16:30 Productization of Stan Eric Novik. Hall TU1, 1st floor.

16:30 - 17:30 Model assessment and selection Aki Vehtari. Hall AS2, 2nd floor.

19:00 - 21:30 Get together, Musiikkitalo, Mannerheimintie 13 A, Helsinki

Thursday 30th August

Location: Aalto University, Töölö campus, Runeberginkatu 14-16, Helsinki. Nokia hall.

08:45 - 09:15 Welcome

09:15 - 10:00 Invited Talk: Hierarchical modelling of galaxy clusters for Cosmology Maggie Lieu, European Space Astronomy Centre

10:00 - 10:45 Coffee break

10:45 - 11:45 Contributed Talks: Session 1
Session Chair: Richard McElreath

Arya A. Pourzanjani, Tie Bo Wu, Benjamin B. Bales, Linda R. Petzold Relating Disparate Measures of Coagulapathy Using Unorthodox Data: A Hybrid Mechanistic-Statistical Approach
Martin Modrák ODE Model of Gene Regulation
Sebastian Weber and Eero Siivola Solving ODEs in the wild: Scalable pharmacometrics with Stan

11:45 - 12:00 Poster Spotlights

12:00 - 13:00 Lunch

13:00 - 14:00 Contributed Talks: Session 2
Session Chair: Theresa Smith

Iiris Sundin Using counterfactual queries to improve models for decision-support
Alex Braylan and Daniel Marthaler Self-tuning Holidays
Jari Turkia Modeling the Effects of Nutrition with Mixed-Effect Bayesian Network

14:00 - 14:45 Invited Talk: Bad Data, Big Models, and Statistical Methods for Studying Evolution Richard McElreath, Max Planck Institute for Evolutionary Anthropology

14:45 - 15:00 Poster Spotlights

15:00 - 15:45 Coffee Break

15:45 - 16:45 Contributed Talks: Session 3
Session Chair: Ilaria Prosdocimi

Paul Buerkner Define Custom Response Distributions with brms
Kristian Brock Clinical trial designs in Stan with trialr
Marco Munda Analysis of repeated measures data in RStan

16:45 - 17:15 Stan Roadmap Bob Carpenter, Columbia University

17:15 - 18:00 Networking

19:30 - Dinner

Friday 31st August

Location: Aalto University, Töölö campus, Runeberginkatu 14-16, Helsinki. Nokia hall.

09:00 - 09:45 Identifying the effect of public holidays on daily demand for gas Sarah Heaps, Newcastle University

09:45 - 10:00 Poster Spotlights

10:00 - 10:45 Coffee Break

10:45 - 11:45 Contributed Talks: Session 4
Session Chair: Maggie Lieu

Bruno Nicenboim The implementation of a model of choice: the (truncated) linear ballistic accumulator.
Ville Laitinen & Leo Lahti A Hierarchical extension to Ornstein-Uhlenbeck-type Student’s t-processes
Christopher Hammill & Jason Lerch Analyzing Brain Structure Trees with Stan

11:45 - 12:00 Poster Spotlights

12:00 - 13:00 Lunch

13:00 - 14:00 Contributed Talks: Session 5
Session Chair: Sarah Heaps

Jonathan Auerbach, Timothy Jones, and Robin Winstanley Predicting New York City School Enrollment
Leo Egidi, F. Pauli & N. Torelli Are shots predictive of soccer results?
Gertjan Verhoeven Does accounting for an artificial turf advantage in Dutch football increase predictive accuracy of probabilistic models?

14:00 - 15:00 Contributed Talks Session 6
Session Chair: …

Thomas A. Nielsen, Dominic Steinitz & Henrik Nilsson Getting more out of Stan: some ideas from the Haskell bindings
Yuling Yao, Aki Vehtari, Daniel Simpson and Andrew Gelman PSIS-diagnostics of ADVI in Stan
Rok Češnovar, Davor Sluga, Jure Demšar, Steve Bronder, Erik Štrumbelj GPU Optimized Math Routines in the Stan Math Library

15:00 - 15:45 Coffee Break

15:45 - 16:15 Stan Interface Roadmap Jonah Gabry, Columbia University

16:15 - 17:00 Esther Williams in the Harold Holt Memorial Swimming Pool: Some thoughts on complexity Daniel Simpson, University of Toronto

Posters

Shrinkage priors for Bayesian penalized regression
van Erp, S., Oberski, D. L. & Mulder, J.
Kinetic modelling of cocoa bean fermentation: An application of Bayesian parameter estimation using RStan
Moreno-Zambrano, M., Grimbs, S., Ullrich M. S., & Hütt, M.
Hierarchical Bayes goes marine: building propulsion power prediction models with Stan
Solonen, A., Laine, M. & Haario, H.
Who drinks in South Africa, and how much? Integrating multiple sources of information to estimate alcohol consumption trends
Cois, A.
Estimating flood probability Bands Using Flood Event Data - a case study on the importance of specifying prior distributions
Prosdocimi, I.
Relating the development of maternal Post-Traumatic Stress to Health Related Quality of Life of children with severe burn injuries
Veen, D., Egberts, M.R., Veen, D., van de Schoot, R. & van Loey, N.E.E.
Bayesian inference with partial differential equations using Stan and adjoint sensitivity analysis
Zhang, Y. & Gillespie, W. R.
Hierarchical Gaussian modelling of instantaneous heart rate distributions
Korpela, J.
Using counterfactual queries to improve models for decision-support
Sundin, I., Schulam, P., Siivola, E., Vehtari, A., Saria, S. & Kaski, S.
Hierarchical Bayesian approach for modelling region- and event-level variability in flood damaging processes
Sairam, N., Schröter, K. & Kreibich, H.
Can predicting data improve model interpretation and inferences? Using Stan to fight the assumption of independence.
Bartoš, F., Lukavský,J. & Děchtěrenko, F.
ctsem – hierarchical continuous time state space modelling for R
Driver, C.
Modelling the effect of billboard advertising on online sales with Rstan
Groß, M.
V-Dem: Aggregating Expert Opinion on Democracy with IRT Models
V-Dem Institute
Type M error in practice: A case study
Vasishth, S., Mertzen, D. Jäger L., Gelman, A.
Age period cohort models with unequal time intervals
Smith, T.
Qualifying drug dosing regimens in pediatrics using Gaussian Processes
Siivola, E. Weber, S. & Vehtari, A.
Modelling in vivo kinetics of human T stem cell memory cells
Mackerodt, J., Rubina, A., Miners, K. L., Ahmed, R., Price, D. A., Baird, D. M., Macallan, D., Ladell, K., Asquith, B. Bayesian inference of protein conformational ensembles from limited structural data
Potrzebowski, W.
Using STAN to Explore Fairness in University Admission Policies
Mathioudakis, M., Castillo, C.
A Trial By Fire: Presenting A Prospective Bayesian Analysis Plan To a Room of Statisticians
Clark, O.
Bayesian model for the destruction of bacteria and bacterial spores under heat treatments
Ilvonen, L., Honkela, A.
Bayesian Regularized Mediation Analysis
van Kesteren, E. Oberski, D.
Forecasting a Stock’s Remaining Intraday Volume
Bari, O. Constandache, A.
Sequential Parameter Learning of Markov Switching Stochastic Volatility Model with Fat-tail Innovations
Yin, M.
Estimating individual response to treatment
Winkelbeiner, S. & Homan, P.
Stan Governing Body
Baldwin, B.

Get together

Get together event will be on Wednesday evening at Musiikkitalo, Mannerheimintie 13 A, Helsinki (included in the registration fee)

Dinner

Conference dinner will be on Thursday evening (included in the registration fee).

Tutorials

Tutorial videos

Pre-conference tutorials led by Stan developers are organized for 29th August 2018 at Aalto University Otaniemi campus, Maarintie 8, Espoo, 13mins metro ride away from Helsinki city center. There are two parallel sessions. The tutorial sessions are:

Basics of Bayesian inference and Stan (morning half day), Jonah Gabry & Lauren Kennedy
In this tutorial we will first review some of the foundational concepts in Bayesian statistics that are essential background for anyone interested in using Bayesian methods in practice. Then we will introduce the Stan language and the recommended workflow for applied Bayesian data analysis by working through an example analysis together. Since we only have 1/2 day for this tutorial it will be beneficial for participants to have at least some previous experience with statistical modeling, but prior experience fitting Bayesian models is not a requirement.
This session will be helpful preparation for anyone interested in attending the afternoon session on hierarchical modeling. We will be interfacing with Stan from R, but users of Python and other languages/platforms can still benefit from the tutorial as all of the code we write in the Stan language (and all of the modeling techniques and concepts covered) can be used with any of the Stan interfaces.
Hierarchical models (afternoon half day), Ben Goodrich
Hierarchical models are and should be the default way to conduct a Bayesian analysis. By hierarchical Bayesian models, we mean any generative process where the distribution for some unknown depends on one or more other unknowns. This includes instrumental variables, interaction terms, hierarchical shrinkage priors and many others. However, the most common instance of a hierarchical model is one where the generative process allows some of the parameters to vary by group, which can be conveniently specified using the syntax of the lme4 R package. In this tutorial we will start with estimating and interpreting estimates of hierarchical models using the stan_glmer function in the rstanarm R package, proceed to the brm function in the brms R package, and finish with some examples where participants modify or write code in the Stan language to specify a hierarchical model.
Prerequisites: Some experience with R (not necessarily the aforementioned packages) and some experience with MCMC (not necessarily Stan). Slides
Stan C++ development: adding a new function to Stan (morning half day), Bob Carpenter, Sean Talts, and Mitzi Morris.
This tutorial will prepare you to submit your first pull request for Stan’s C++ libraries. We will cover all of the steps required to add a new function with analytic gradients to the Stan language, including creating the issue, branching in GitHub, code organization and style, unit testing, continuous integration, code review, API and user-facing documentation, and merging. In addition to the process, we will survey the math library, including scalar and matrix data types, automatic differentiation, and template traits and metaprograms.
Ordinary differential equation (ODE) models in Stan (afternoon 1.5 hours), Daniel Lee
This tutorial will cover the usage of the ordinary differential equation (ODE) solvers from within the Stan language. It will be a hands-on session where we’ll learn to specify ODEs as functions, the use of the ODE integrators, and practical tips on packing and unpacking the arguments to the function.
This tutorial is meant to be hands-on. Please be familiar with running Stan models from the interface of your choice. It will help to know the difference between arrays of real and vector types (see Ch. 26 “Matrices, Vectors, and Arrays”).
Productization of Stan (afternoon 1.5 hours), Eric Novik
We all love Stan for its expressivity and flexibility, and for making state-of-the-art NUTS sampler available to a broad community of statisticians, econometricians, pharmacometricians, data scientists, and many other quantitative analysts.
What we don’t love is munging and mapping data into Stan, slowly iterating over many models, and deploying models into production for use by non-specialists. Integrating a robust Bayesian workflow from research into production is not trivial but it is extremely rewarding as it puts the product of our work in the hands of people who need it most.
Many organizations have started doing this with Stan and there are some patterns of use as well as anti-patterns that are emerging as a result. In this session, we will present several approaches to productizing Stan and discuss their strengths and weaknesses.
Productization panel discussion: Markus Ojala (Smartly), Tom Nielsen (Tweag.io), Anna Kircher (Lendable), Eric Novik (Generable)
Model assessment and selection (afternoon 1 hour), Aki Vehtari
In this tutorial I will review estimation of predictive performance of models in M-open seetting where we assume that none of the models is the true model. When we don’t trust our models, we can re-use data as a model-free proxy for the future data distribution. To adjust for data re-use we use leave-one-out cross-validation for exchangeable observations, K-fold cross-validation for group exchangeable observations, and m-step ahead validation for time series. I present how these performance estimates and related predictive distributions can be used for model comparison, averaging and checking. I discuss fast importance sampling approximation including useful diagnostics. Finally I discuss relations to alternative methods for model selection and weighting.
Tutorial demonstrations are based on R with rstan, rstanarm, brms and loo packages.