ENBIS-15 Pre-Conference Workshop on "Generalised Regression – A Silver Bullet for Statistical Modelling?"

6 September 2015; 13:00 – 17:00; Czech Technical University, School of Mechanical Engineering, Prague (Czech Republic)

This half-day ENBIS-15 pre-conference workshop will be given by Volker Kraft (JMP Academic Ambassador, Europe).

Do your analyses using stepwise regression or other mainstream techniques sometimes yield unsatisfactory results? Do your models often over-fit and generalize poorly to new data? How do you decide which variables to remove from your model, and how much time do you lose manually preprocessing data so that restrictive modeling assumptions are more likely to hold? Or are you a confident user of Generalized Regression already, but curious to see the version 12 enhancements in JMP Pro, like informative diagnostics features, interactive model selection or inverse prediction? Interesting questions? Then this workshop is right for you.

Generalized Regression models use penalization techniques and often work well, even with challenging data. Generalized Regression can handle non-normal responses, deal with correlated predictors and high-dimensional data (with more predictors than observations), and perform variable selection and shrinkage - all within a unified framework. The penalization techniques available within JMP Pro include Ridge, Lasso and Elastic Net. Effectively harnessing these relatively new techniques is as easy as using any other modeling personality in the Fit Model platform – simply identify your response, construct model effects and pick the desired estimation and validation method. JMP Pro automatically fits your data, performs variable selection when appropriate, and builds a predictive model that will generalize well to new data. The Generalized Regression personality also gives options to choose the most appropriate distribution for the response you are modeling.

As well as introducing the specific techniques used to perform Generalized Regression, the workshop introduces the concepts of bias-variance tradeoff, cross-validation and model evaluation. Multiple data sets will be analyzed to exemplify a variety of modeling challenges and use cases. Experience with JMP Pro is not mandatory.

Participants are invited to bring their laptops (Windows or Mac) into the class. JMP Pro 12 can be temporarily installed free of charge to work through the examples and exercises in the workshop.

The pre-conference workshops will take place at the Czech Technical University, located at the Charles Square, in room 104.