R and Shiny lightning talks at Hymans Robertson

Hymans Robertson is excited to host three 20-min lightning talks on R and R-Shiny. The event is over lunch on June 2nd in our Edinburgh office, and we will be providing some refreshments to attendees, as well as plenty of conversation/networking.

We have three presenters.

Presentation Title: Weighted Monte Carlo. A technique for post-processing "calibration".

Presenter: Louise Juul Mousten

Abstract: When calibrating models such as Economic Scenario Generator (ESG) models, you are often faced with the challenge of choosing between speed of calibration and fit of models. Hence a choice between simple models with few parameters and more complex models with the flexibility to match distributional targets.

An alternative approach is a post-processing modification of simulations by reweighting instead of assuming equally weighted simulations. Reweighting is a flexible technique for matching calibration targets or imposing alternative views, no matter the complexity of the underlying models used to generate the simulations.

This talk will introduce maximum entropy as a method for reweighting of Monte Carlo simulations, together with reweightR, an R package for producing and evaluating weights used for exploring post-process scenarios.

Presenter Bio: Louise Juul Mousten is a Risk and Modelling Consultant within the Insights and Analytics team, at Hymans Robertson. She is involved in modelling and calibration of the ESS models, specifically leading on the R/C# projects in creating prototype models and other bespoke modelling project such as the reweightR. Prior to joining Hymans, she has 5 years' experience working with risk management and quantitative investment strategies, primarily focused on modelling solvency capital requirements and incorporation of economic scenario generating in asset/liabilities modelling within the pension industry in Denmark.

Presentation Title: When to carry an umb-R-ella: decision-making under forecast uncertainty

Presenter: Dr Jethro Browell

Abstract: Probabilistic forecasts quantify uncertainty and enable improved decision-making in many settings but are much more challenging to work with than more familiar point forecasts. When faced with asymmetric penalties for forecast errors, or when our aim is to control risk, probabilistic forecasts are a necessary to make “optimal” decisions under uncertainty. This talk will introduce probabilistic forecasting and ProbCast, an R package for producing, manipulating, and evaluating predictive distributions. Also showcasing the library via a demo app.

Presenter Bio: Jethro Browell is a Senior Lecturer and recent EPSRC Innovation Fellow at the University of Glasgow’s School of Mathematics and Statistics. His interests span forecasting methodology and analytics that empower decision-makers with uncertainty information allowing them to make better decisions and control risk. Jethro has worked extensively with the energy industry in the UK and Europe developing methods for modelling and forecasting renewable power production, electricity demand, and other quantities of interest. Several electricity generators and network companies are using his forecasts and decision-support tools today.

Jethro is actively involved with the IEEE and International Energy Agency as an editor and working group leader and is currently seconded into National Grid ESO’s forecasting team on a part-time basis. His publications and other outputs can be found at [www.jethrobrowell.com](https://eur03.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.jethrobrowell.com%2F&data=05%7C01%7CPaul.Hammant%40hymans.co.uk%7C4191b0d7c1c549c0057508db368fa4bf%7Ca2276d23b28149629c993c5c8d9895c5%7C0%7C0%7C638163762792788648%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=BEpZSZyuccBna8ctrLFIf%2F6wvRqxJ4fYBIzsSoGaJ0I%3D&reserved=0).

Presentation Title: Lexis surfaces: Revealing the hidden landscapes of risk by age and time

Presenter: Dr Jon Minton

Abstract: Much as a spatial map shows elevation by latitude and longitude, Lexis surfaces show how risks and other population attributes vary by age and time. This presentation will (re)introduce the concept of the Lexis surface as it relates to more common mortality-related measures, present a number of use-cases from public health and the social sciences, then show how Lexis surfaces can be generated and interacted with in dashboards using the HMDHFDplus, shiny, and rplotly packages.

Presenter Bio: Dr Jon Minton is a public health intelligence adviser based in Public Health Sciences in Public Health Scotland, currently also learning to be a software developer at Codeclan. He has a keen and long-standing interest in demographic visualisation, and been an R user for around twenty years. He has worked as a systematic reviewer and health economist and spent too long at university.