Why I switched to ModelRisk and why I wrote my book
By Dale Lehman, Alaska Pacific University
I have been using Monte Carlo spreadsheet simulation software since 1989 in both my consulting work and in courses I teach. I liked the Crystal Ball software I was using but was growing disenchanted with the increasingly difficult licensing issues and lack of development that arose with each acquisition of the company that made the software. I became aware of ModelRisk several years ago, and was intrigued by some of its advanced capabilities.
There was a problem – I needed a simple, focused book to teach from. The software documentation was robust, and David Vose has written a very comprehensive text on risk analysis modeling, but both were too detailed for the type of students I teach (MBA students, mostly). I asked the folks at Vose if they could support my writing a text and they agreed. So, with Huybert’s and Greg’s help, that is what we did.
Our book has a number of distinguishing features. It uses real data and interesting issues (e.g., modeling the cost of health insurance plans, the variability of financial returns for making Hollywood movies, the probability of extinction of a fish species, estimated total seasonal snowfall at a ski resort, etc.). Every chapter contains exercises from 8 different industry focus areas. Since ModelRisk has a number of unique capabilities, the book provides examples that highlight these – here are a couple of examples:
Other packages include the capability to correlate distributions in a simulation model, but none is as robust in the copulas that are available. This capability comes at some cost – it is a bit more cumbersome to enter the correlations than in some other packages. So, we needed an example to show why it is worth this extra complexity. We found it in house price data from the financial meltdown. Home prices in different parts of the US were best not modeled by a linear correlation. The Clayton copula fit the data quite nicely and captured an essential feature of the crisis: home prices were more strongly correlated when prices were falling than when they were rising.
Parameter and Model Uncertainty
When dealing with sparse data, the best fitting distribution should be viewed with some suspicion. ModelRisk permits the best fitting distribution to reflect the fact that the parameters of the distribution are uncertain. Bayesian Model Averaging is available to reflect the fact that we are unsure which of the candidate distributions to use (the best fitting one may not be that much better than the 2nd or 3rd best). One application we provide is to estimate the probability of very low birth weight babies in relation to whether or not the mother was a smoker – the odds ratio is quite sensitive to the choice of distribution and the degree to which we recognize uncertainties in the fit .
The book also covers Markov Chains, expert opinion modeling, regression models, GARCH time series models, frequency-severity modeling, and optimization. A number of new tools have been added to ModelRisk since publication, including differential equations, new distributions, best fitting auto-selection, etc. and they will certainly be in the next edition!
I’ve used the book in several classes now. It is a challenging book, as the exercises are real and require careful modeling. Students are certainly engaged and the course already has a reputation as the most challenging (despite the lack of mathematics in the text) in our MBA program. The amount of effort the students have devoted to these classes is evidence both of their dedication to their education, as well as the ability of the book to capture and sustain their interest.
More on Dale Lehman and the book "Practical Spreadsheet Risk Modeling for Management"
- For specific questions on the textbook please contact Dale Lehman.
- Further details on Dale Lehman(eg. complete vitae) can be found here.
- For more information on Alaska Pacific University, please visit their website here
Free Webcast: "How to Teach Risk Analysis"
On December 10, 2012 both Dale Lehman, author of "Practical Spreadsheet Risk Modeling for Management" and professor at Alaska Pacific University, and David Vose, author of "Risk Analysis: A Quantitative Guide" and teacher at various inhouse and onsite training courses and universities like Harvard, will explain the best practices in teaching risk analysis. If you are a professor or a trainer this one hour webinar is perfectly suited to your needs.
This webinar is part of our weekly series of free webinars. The topics for the following weeks are displayed here.
|How to Teach Risk Analysis
||Dale Lehman / David Vose
||Dec. 10, 2012
||11AM EST / 4PM UTC
We Are Here For All Your Risk Analysis Requirements:
Software, Training & Consulting