Workshop Leader:Tom Keelin, Managing Partner, Keelin Reed Partners

Tom Keelin has combined a career in decision analysis practice with innovations to advance the field. As Chairman of Millennial Capital, LLC, he has served as general partner for multiple successful real estate funds. He leads strategic decision-making for acquisitions, operations, dispositions, and portfolio management – using decision-analysis modeling and simulation. Tom is also a founder and Managing Partner of Keelin Reeds Partners, a consulting firm that has helped over a hundred venture-funded and mid-market life sciences companies with strategy and business development. In both roles, he recognized the need for better continuous-uncertainty representations to support decision-making. Addressing this need, Tom invented the metalog distributions, recently published in the DA Journal.

Previously, Tom was Worldwide Managing Director of the Strategic Decision Group, head of its London office, and founder of its life sciences practice. Tom supervised the client work for and co-authored the Harvard Business Review article “How SmithKline Beecham Makes Better Resource Allocation Decisions” (Mar-Apr ’98). Through that work, he and his colleagues invented the portfolio-management standard which subsequently was adopted widely across life-sciences industry. Earlier, with Decision Focus, Inc, Tom developed the Over/Under Capacity Planning model, which was recognized as one of EPRI’s top ten all-time contributions to the US electric power industry.

Tom is a Fellow of the Society of Decision Professionals and a founder and director of the Decision Education Foundation. Tom holds three degrees from Stanford University: BA in Economics and MS and PhD in Engineering-Economic Systems.

Overview. The metalog distributions are the most flexible probability distributions in the world. They are also among the easiest to use. Come to this workshop to equip yourself and your teams with understanding and tools for practical use. Metalogs make it disarmingly simple to model expert assessments with a smooth continuous distribution (both CDF and PDF), simulate it, and gain otherwise-unavailable insights from simulated or empirical data.

The metalog distributions are new. They displace 100-year-old family of Pearson distributions (Normal, Beta, Student t, Gamma, Chi Squared, F, and others) by being more flexible, easier to use, easier to interpret, and faster to simulate. In contrast to the Pearson distributions, the metalogs are flexible enough to represent virtually any distribution arbitrarily closely; allow users to choose among unbounded, semi-bounded, and bounded forms; are easy to fit to data using ordinary least squares (displacing difficult non-linear parameter estimation); have simple closed form equations; and enable Bayesian inference in closed form.

Agenda. The workshop will provide an overall introduction to metalogs. Math will be kept to a minimum – just enough to enable intuition of what metalogs are and why and how they work. The emphasis will be on properties and tools that are useful in practice. We will show how fit metalogs to a wide range of data sets (expert assessment, simulation results, and empirical data) and discuss insights that emerge. We will spend focused time on how to deal with challenges including infeasible data sets, overfitting, and choosing the right number of terms. Regularization and visual-inspection techniques will be introduced to help overcome these challenges and applied to real data sets. We will introduce metalog regression and Bayesian inference: updating an “any shape” prior probability distribution with new data to produce an “any shape” posterior in closed form according to Bayes Theorem (prior to metalogs, this has never been possible).

Who Should Attend? Practitioners – including supervisors, leaders and team members who use probabilities in practice. Professors and teachers – anyone who teaches probability and its practical applications in universities, corporations, or elsewhere – in engineering, computing, machine learning, management, medicine, statistics, decision analysis or any other field.

Why Should You Attend? Metalogs can displace all other continuous distributions you currently use. They are more flexible and simply better. Metalogs are being increasingly widely adopted. They have been endorsed by Professor Ron Howard as “A Tour de Force” and Nobel laureate Harry Markowitz with “I believe!”. This is your opportunity to learn directly from the inventor.

Bring Your Laptop and Data Sets. Bring your laptop equipped with Excel. Excel files that implement the metalog system will be provided. You will be encouraged to download this software (including an optional Excel add-in enabling metalog simulation) prior to the workshop. Also bring data sets typical of your applications. You will be encouraged to send these to Tom in advance for use as appropriate in workshop demonstrations.