|
|
|
|
  |
| |
|
|
Duration: 5 days
Course overview
This advanced course provides an in-depth knowledge of the modeling techniques necessary for international level risk assessments in Epidemiology. We critically look at risk models, and the participants are encouraged to bring along modeling problems they are currently faced with. Previous attendance to our Quantitative Risk Analysis and Disease Modeling in Epidemiology course, or to module 1 of the Animal Agriculture and Food Safety Risk Analysis course is required. If you have not attended to either course but have experience in risk analysis, please contact Dr. Zagmutt (course instructor) to make sure you have the knowledge necessary to attend to the course.
This course is meant to be the second of two modules on quantitative risk assessment, so participants wanting to get comprehensive training in the subject are encouraged to take it in conjunction with the Quantitative Risk Analysis and disease modeling in Epidemiology course that we offer the week immediately prior to this course. Please contact us for special pricing available when taking both courses together.
Although similar to Module 2 of Animal Agriculture and Food Safety Risk Analysis, this course focuses solely on risk analysis methods applied to epidemiology, and doesn’t explicitly cover specific food safety modeling techniques.
This module is suited to those already familiar with a modeling environment, who have some modeling experience and who are interested in developing these abilities further. The module content will enable the participants to produce realistic, professional quality models. It is designed to encourage the modeler to develop creative problem solving skills through plenty of problem exercises.
Please review the level of computer (Excel and Windows) knowledge necessary before attending the course. The requirements are very basic, but ensuring that all participants arrive with this basic level prevents us from wasting too much time on familiarization with Excel rather than learning about risk modeling.
top |
Training material
All lecture notes are provided as PowerPoint files. A CD of these files is provided to each participant. Printed handouts are also provided. The CD also contains all model files produced for the course and extra references and documents of interest. Any extra models developed during the course are downloadable from a private website dedicated to the course.
top |
Course format
The course runs from 09:00 to 17:00 each day. Lunch, and morning and afternoon coffee are provided. A group dinner on Thursday night is also included.
top |
Who should attend
Professionals working in risk assessments in animal health, human health, food safety, and plant pathology, and risk managers who have some basic knowledge of spreadsheets and simulation modeling. The course is also well suited for scientists providing input to a risk assessment.
top |
Prerequisites
Participants will be primarily using ModelRisk (version 3.0 or newer) with Microsoft Excel to solve the exercises. It is essential that all participants are reasonably proficient in Excel (see prerequisites). Also, the R statistical language and WinBugs will be used for a few sections. However, participants can also choose to use another Monte Carlo package of their choice (@Risk, CrystalBall, etc). We will provide participants with a CD with the installer files for ModelRisk, R, and WinBugs, but participants should come to the class with all the required software installed. Participants can find trial versions of ModelRisk here http://www.vosesoftware.com/trial.php, the R installer here http://cran.r-project.org/, and WinBugs's installer here http://mathstat.helsinki.fi/openbugs/OpenBUGS.zip. As the trial license of ModelRisk lasts 30 days, participants should install it only a few days before the course.
In addition to ModelRisk, R and WinBugs, participants are required to bring laptops loaded with Microsoft Word, PowerPoint, and Excel, and with a functional CD drive.
For @Risk users, trial copies are available free of charge from Palisade's web-site but it should not be installed too early as trial versions run out after 10 days.
Participants very proficient in the R statistical language can develop the exercises using this platform. The instructor will help them solve their problems, but the solutions to group problems will be shown in Excel and ModelRisk.
top |
Free risk analysis reference
The ModelRisk help file is a comprehensive risk analysis training reference and is free of cost. This help file provides an in-depth explanation of all of the risk analysis concepts, techniques and methods introduced in this course and greatly complements the course material. The help file is included in the installation file of ModelRisk and remains active even after the demo version expires. It can also be accessed directly from Vose Software's website here
top |
Teaching philosophy
All of Vose Consulting's courses aim to help the participants understand (rather than 'learn') risk analysis, which can only be achieved through a relaxed, informal and interactive environment, through plenty of examples and hands-on exercises where students apply and adapt what they have learned.
top |
Course content
Day 1
- Review/refresher of material of introductory courses:
- Probability theory
- Stochastic processes
- Binomial process
- Poisson process
- Hypergeometric process
- Central Limit Theory
- Mixed Poisson and Binomial processes
- Problems to solve
- Uncertainty and variability
- Meaning of uncertainty and variability, the value of their distinction, modeling techniques
- Examples of modeling problems where they are usefully separated
- Structure of two-dimensional (second order) risk analysis models
Day 2
- Classical statistics - applications for uncertainty analysis
- Estimation of population mean and standard deviation
- Estimation of population prevalence and Poisson mean
- Bayesian inference for uncertainty analysis
- Theory and derivation
- “Manual” construction of a Bayesian model, and simulation results
- Problems to solve
Note: Bayesian methods of statistics is used here considerably as an intuitive means of helping participants understand the connection between data and knowledge
Day 3
- Bayesian statistics (cont.)
- MCMC/WinBUGS modeling
- Latent-class analysis: diagnostic testing in absence of gold standard
- The Bootstrap - applications for uncertainty analysis
- Non-parametric and parametric Bootstrap techniques
- Use of Jack-knife for gauging robustness
- Applications and problems to solve
Day 4
- Modeling dependencies
- Rank order correlation, conditional logic and indexing, envelope method, copulas, bootstrap
- Problems to solve
- Analyzing and using data
- Checking quality and appropriateness
- How to accept and reject different data sets
- Spotting the traps and filling the gaps in reported data
- Determining distributions from data
- Assessing validity of data
- First order distribution fitting
- Fitting to parametric and non-parametric distributions
- MLE and goodness of fit statistics
- Using linear solvers with gof statistics
- Second order distribution fitting
- Parametric and non-parametric distributions
- Likelihood estimating, Bootstrapping, other methods
- Problems to solve
Day 5
- Critiquing and reviewing risk analysis models and results
- Review of past models
- Finding common modeling mistakes, and how to prevent them
- Presenting risk analysis results
- Statistical and graphical outputs from a risk assessment
- Discussion of modeling problems from course participants
top |
Dr. Francisco Zagmutt has an impressive command of the tools used in conducting a quantitative risk analysis which include frequentist and Bayesian probability theories, statistics, mathematics, and programming in Excel, R, WinBugs and ModelRisk, that go well beyond his formal education. He offered a wide array of references with a degree of technical difficulty that reflects his strong interest in the subject matter and his commitment to stay current. In addition to an impressive knowledge base, Francisco is an excellent teacher. The class contained individuals with diverse backgrounds from professional to academic degrees and he brought us all along with him. He was able to answer all questions and provide hands-on exercises that were relevant to conducting a QRA. I highly recommend the class.
Vicki Lancaster, Ph.D.
Mathematical Statistical FDA/CVM |
|
|
|
| |
|
|
|
| | |
©Copyright 1997-2010 Vose Consulting. All Rights Reserved.
|
|
|
|
|
|
