Introduction to MLM using MLwiN

Course Content:
This course explores the use of multilevel models and emphasises their practical application in social sciences.  The main focus is on multilevel models for continuous outcomes, with binary outcomes also covered at an introductory level.  Analyses are illustrated using MLwiN (a package dedicated to multilevel modelling available free to academics).

UK academics ONLY are entitled to apply for a free MLwiN download thanks to the LEMMA project funded by the ESRC's National Centre for Research Methods(NCRM).

Course Objectives:

  • To refresh participants' knowledge of using single-level regression
  • To teach the principles and assumptions underlying multilevel models for analysing continuous and binary outcomes.
  • To introduce participants to the use of multilevel techniques.
  • To provide examples of analysing data using the multilevel package MLwiN and interpreting the output.

The topics covered in this course include:

  • Introduction to Multilevel Models
    • Recap on regression
    • When are Multilevel Models Appropriate?
    • How does Multilevel Modelling Differ from Regression?
    • Why Use Multilevel Models?
    • Historical perspective
    • Examples
    • Regression models can be misleading!
    • Model Specification
    • Non-normal data
    • Multilevel Terminology
  • Introduction to MLwiN
  • Multilevel Models with Two Levels
    • Variance Components Models
    • Random Intercept Models
    • Adding Explanatory Terms
    • Continuous Explanatory Variable
    • Two Explanatory Variables
    • Interactive Effect of Two Explanatory Variables
    • Explanatory Variables Measured at the Group Level
    • Cross-level Interactive Effect
    • Analysing a Subset of the Data
  • Random Slope Models
    • Visualising the Random Slope Model
    • Model Definition
    • Schools Example
    • Interpreting Level 2 Variances
    • Interpreting the Level 2 Covariance
    • Testing the Significance of Random Slopes
    • Models with Several Random Slopes
    • Residuals
    • Fitting random slope models in MLwiN
  • Longitudinal Data
    • Longitudinal Studies and Multilevel Data
    • Example - The Edinburgh Study of Youth Transitions and Crime
  • Other Data Structures
    • Do you need a Multilevel Model?
    • Alternative Multilevel Models
    • Combining Different Structures – Matching the Complexities of Real Life
    • When Not to Use Levels
  • Multilevel Models for Binary and Binomial Data
    • The Binary Model
    • Single-level Model
    • Two-level Model
    • Explanatory Variables
    • Binomial Data
    • Model definition
    • Practical Application and Interpretation
    • Examples

The course manual also includes sections on:

  • General Points on Fitting Multilevel Models
    • Defining Random Effects: When is a variable a level?
    • Significance Testing
  • Model Fitting Methods
    • Maximum Likelihood Methods
    • Methods Similar to Maximum Likelihood
    • Simulation methods
    • Method choice
  • Models for Multinomial Data
    • Multilevel Multinomial Models
    • Example – British Election Study
    • Extending a Binary to a Multinomial Model
    • Moving to More Than Two Response Categories
    • Notation
    • Interpretation - Odds Ratios
    • Interpretation - Probabilities
    • Multilevel Multinomial Models
    • Variance of Level 2 Random Effects
    • Multilevel Model of Voting Behaviour
    • Between Constituency Variation in Odds Ratios
    • Constituency Differences in Probabilities
  • Ordinal and Count Data
    • Ordered Multinomial Data: Single-level Model
    • Ordered Multinomial Data: Two-level Model
    • Count Data: Single-level Model
    • Count Data: Multilevel Model

Request the training materials

View all courses available