Exploratory Data Mining via SEARCH Strategies


This workshop provides an overview of current techniques in exploratory data mining for quantitative research in the social and behavioral sciences. Exploratory data mining uses computational methods on the large amounts of data that are now available in the social and behavioral sciences in order to construct predictive models of behavior. It assesses the predictive value of all variables, or combinations of them, within a given data set instead of the standard hypothesis testing of many standard statistical techniques. These data mining techniques can be used to model categorical choices, to classify groups, to discover patterns, and to model longitudinal data. Exploratory data mining techniques can be fruitful in most situations where categorical regression or many multivariate analytic techniques are used.

This workshop will explore key algorithms including regression trees and SEM models (CART, SEMtrees, PARTY, etc.). This work was initiated by the SAS algorithm SEARCH (Morgan & Sonnquest, 1963), and the workshop will begin here and then move to use of the free software modules in R that are currently used for exploratory data mining. The workshop will offer a mixture of mathematical statistics in the morning session and practical, hands-on work in the afternoon. Participants are encouraged to bring their own data to which they can learn to fit an exploratory model and to write up the activity.

Background: Accepted applicants should have a strong background in regression, and have some familiarity (although not necessarily proficiency) with programming in a statistics package. The course will use R and individuals coming from only a "point and click" background may find this part challenging. It would be even more helpful, but not necessary, if individuals had some knowledge of multivariate statistics or structural equation modeling.

Fee: There are no tuition fees for accepted participants.

Application: Admission is competitive and seating is limited. Apply using the Summer Program Portal (by clicking on the "Registration" tab at the top of this page) to provide your information and then select the course. Upload the following documents via the Portal:

For graduate students:

  • Letter describing reasons for attending course
  • Current CV
  • Unofficial transcript
  • Letter of recommendation from adviser
  • List of classes related to statistics and quantitative methods

For faculty:

  • Letter describing reasons for attending course and research interests
  • Current CV
  • List of courses taught

Deadline for Application: April 30, 2017

Stipends: The first ten (10) admitted workshop participants will receive a travel stipend of $500.

Special Award: Those completing the course will be eligible to compete for two $15,000 awards for innovative uses of SEARCH. For more information and eligibility criteria, please refer to this announcement.

Sponsor: James P. Morgan, Professor Emeritus of Economics, and Research Scientist Emeritus, Survey Research Center, Institute for Social Research, University of Michigan.

Tags: data mining, exploratory analysis

Course Sections