Data Modeling Workshop 05-10-2009

Duration : 5.0 day(s)

:: Course Summary

This 5-day workshop offers participants the opportunity to become familiar with both the most classical statistical modeling techniques – linear and nonlinear regression, logistic regression and its alternatives- and more recent powerful methods – including PLS regression, Lasso models and random forests-. Methods are presented with a strong emphasis on applications and interpretation of results. Participants are encouraged to use their own software and data. Everyday, a special workshop related to a specific aspect of data modeling is offered and ample time is allotted to hands-on exercises.

:: Learning Objectives

Upon completion of this workshop, participants will be able :

To setup an efficient strategy to address model-building situations;

To fit a variety of statistical models adapted to the data at hand;

To identify issues that may affect the validity of the results and find sound solutions;

To evaluate the quality of a fitted model, whether in terms of goodness of fit or predictive ability;

To compare different models and to select the most appropriate depending on the context;

:: Target Audience

As this is the case for most sessions offered by Creascience, the mathematical content of the session is kept at a minimum level to focus instead on practical issues, interpretation of software output and usage of the results. Therefore, it can be of interest both for non-statisticians and statisticians, who deal with modeling problems, want to improve their practices and become familiar with more recent techniques. Researchers, scientists, business and finance analysts, engineers and data analysts constitute the typical audience of these classes.

:: Prerequisite

The only technical prerequisite for this workshop is knowledge of basic statistical tools including descriptive statistic, classical plots like scatter plots and histograms and the principle of hypothesis testing (see Fundamental Tools in Statistics or an equivalent course). Prior knowledge of linear regression is not mandatory, but an understanding of the purposes of model building can definitely be helpful in making the most out of the session.

It is imperative that participants are familiar with the way to carry out data handling/manipulation in their statistical software.

:: Topics Covered

Day 1

General introduction to data modeling
Simple and multiple linear regression
Special Workshop: regression graphics

Assessing model quality and goodness of fit
Detecting outliers and influential values
Validating model assumptions

Day 2

Modeling a binary outcomes: logistic regression, discriminant analysis and alternatives
Generalization to any type of discrete outcomes

Ordinal logistic regression
Polytomous logistic regression

Special Workshop: building predictive models
- Why is it different from classical model building?
- Cross-validation and more robust alternatives

Day 3

Dealing with multicollinearity in regression models: a few options

Principal Components Regression
PLS regression
LASSO models
Ridge Regression

Special Workshop: Software packages for data modeling

Day 4

Modern alternatives to data modeling

Regression trees
Random forests

Special Workshop: A general strategy for model building

Day 5

Nonlinear models
A primer on other modeling techniques
Special Workshop: Handling missing values
Extended Data Analysis Workshop
:: Course Content
Instructors
The session is presented by two experienced instructors. They share their practical knowledge of the methods by offering real-life examples and suggesting original applications.
Small class size
The number of participants to this class is limited to 12 people. Combined with the presence of two instructors, this ensures a personalized follow-up of each attendee while maintaining a pace that will suit novice as well as more advanced participants.
Participants
This class offers a unique opportunity to participants to improve their practice and understanding of data modeling in a multidisciplinary environment. Based on our experience, having in the same class participants from various fields working with different data and software packages but sharing similar concerns and goals, allows attendees to get a much wider perspective on the discussed topics as well as tools and solutions available.
Case studies and data analysis
Another specific feature of this class is the time allocated to data analysis. Typically, participants are first asked to reproduce an analysis presented by an instructor. As a second step, they are encouraged to apply the technique discussed to their own data. Additional datasets are also provided in case their data are not suitable for a specific technique. In all cases, emphasis is not limited to analyzing data and interpreting results but first focused on correctly specifying a research question and translating it into statistical terms. All class participants are contacted a few weeks in advance to prepare the workshop and discuss the data they plan to bring to the class.
Software
Participants are encouraged to use their own software during the class. Supported packages include SAS, SPSS, Minitab, Xlstat, S-Plus, R and Statistica. In order to provide attendees with tools for more recent techniques that are not widely available yet, training material will include macros and functions that can fit all types of models discussed during the week.

Upcoming Public Sessions

No public sessions for this training has been planned yet. If you are interested in attending one, please let us know with your prefered training location: Contact us

Current Reviews

There is no review yet for this training.
If you attended it and would like to share your comments with others, follow this link: Write a review
If you would like an overall appreciation of our training courses, you can have a look at the Reviews page of this site.

General Information

Our Philosophy In-House Training Coaching Instructors Latest News Contact Us Newsletter

Tell A Friend