Quantitative Research V: Multilevel modeling

Schedule and location

Wednesday April 19th – Thursday April 20th  

University of Jyväskylä, Jyväskylä, Agora building (visiting address: Mattilanniemi 2), room Lea Pulkkisen sali (4th floor, mornings) and computer class Ag B112.1 (afternoons).

 

Registration 

Registration is open February 20th - April 11th.

Speakers

Dr. Mikko Rönkkö, University of Jyväskylä, Finland.

Organizer

Dr. Mikko Rönkkö, University of Jyväskylä, Finland.

Overview

Many researchers use quantitative research designs and statistical analysis tools without sufficient understanding on what these techniques are based on. This course is a part of a series of six courses where the goal is to develop an understanding of how multivariate statistical methods are used in information systems research and how results are usually presented in journal articles. The main goal of the course is to provide a foundation that enables independent self-study of quantitative research methods. Rather than addressing 

simply how the methods are used, we focus on why certain methods are used and how and why these methods work.

The course is designed for both those interested in just reading and understanding research done with statistical methods and for those who already use or plan to use statistical research methods in their own work.

During the course series we will go through six empirical papers published in leading IS and management journals, and analyze how these papers were done. The methods and research designs used in these papers cover a majority of methods and designs used in these journals. If you master all the methods and techniques presented during the course, you would be ideally capable to publish in these journals – if you have already a new theoretical insight and access to appropriate data.

In this fifth course, we address how longitudinal and clustered data are used in research applications. These data arise from repeated observations of multiple unit, e.g. software development teams, over time, or when the observations belong to groups, for example team members belong to development teams who work in companies. The longitudinal or clustered nature of data presents both opportunities and challenges for researchers. However, these are commonly ignored in information systems research. The course covers both econometrics techniques that aim for eliminating the effects of clustering from the data and multilevel modeling, where clustering is seen as an important feature to be explained.

During the course we address the basics of exploratory analysis of clustered data, specification and estimation of econometric and multilevel models, and interpretation of results. You can complete the course using either Stata, R, or SPSS.

The course has a pre-class readings package and a pre-class assignment that must be returned before the course and a post-class report that must be returned after the course. You should reserve at least two days for the pre-class assignment and at least a day for the post-class report. The course assumes that the participants have a basic understanding of regression models and factor analysis. You can participate the course even if you did not participate on the previous courses.

Detailed Program (preliminary)

 

Day 1 – Morning

Course introduction

Review of courses 1 and 2

Introduction to longitudinal and clustered data

  • Levels in data
  • Between, within, and population average effects
  • Intraclass correlations ICC1 and ICC2

Regression with clustered data

  • Violation of OLS assumptions
  • Different strategies for working with clustered data
  • Centering
  • Interpretation of regression using clustered data

Econometric techniques for clustered and longitudinal data

  • Fixed effects and random effects estimators
  • Using cluster means as controls
  • Hausman test for model comparisons

Day 1 - Afternoon (computer class)

Introducing the data-analysis assignment

Applying econometrics techniques

Day 2 – Morning

Introduction to multilevel models

  • Multilevel regression model
  • Random and fixed effects
  • Random intercept models
  • Random slope models
  • Autocorrelation
  • Error structures in multilevel models
  • ML and REML estimation

Multivariate perspective on longitudinal data using SEM

Day 2 - Afternoon (computer class)

Multilevel modeling using Excel

Applying multilevel modeling

Applying SEM using multivariate perspective

Working on the data-analysis assignment and post-class report

Credit points

Doctoral students participating in the seminar can obtain 2 credit points. This requires participating on both days and returning the pre-class assignment and post-class report.

Registration fee

This seminar is free-of-charge for Inforte.fi member organization's staff and their PhD students. For others the participation fee is 400 €. The participation fee includes access to the event and the event materials. Lunch and dinner are not included.