Introduction to Latent Class Analysis

Session 1: Latent Class Cluster Models
12 Topics | 13 Exercises
A. Basic Ideas of Latent Class Analysis
B. The general probability model for categorical variables
Exercise B.
C. Determining the number of classes/clusters
Exercise C.
D. Fit measures, model specification and selection strategies
Exercise D.
E. Classifying cases into latent class segments
Exercise E.
F. Interpreting LatentGOLD output
Exercise F1.
Exercise F2.
Exercise F3.
G. Example from survey analysis
Exercise G.
H. Including covariates in LC models
Exercise H.
I. Boundary, identification and local solution issues
Exercise I.
J. Extension to continuous variables and other scale types
Exercise J.
K. Including direct effects
Exercise K.
L. Examples with Diabetes data (obtaining scoring equations)
Exercise L.
Session 2: Latent Class Regression Models
3 Topics | 4 Exercises
A. Estimating a Simple LC Regression Model
Exercise A.
B. LC Regression Models with Predictors
C. LC Regression Models with repeated measurements
Exercise C1.
Exercise C2. Example: Latent class growth models
Exercise C3.
Session 4: End of Course Notes
Session 3: Some Advanced Topics
6 Topics | 6 Exercises
A. Incorporating/ accommodating cases with known class membership
Exercise A.
B. Latent Markov modeling for longitudinal data analysis
Exercise B1.
Exercise B2.
C. Multilevel LC models
Exercise C1.
D. Continuous factors (CFactor) and individual-level parameters
E. Controlling for the ‘level’ effect: specifying a random intercept model as an alternative to ‘centering’
Exercise E1.
F. Controlling for the ‘scale’ effect: using scale classes and scale CFactors
Exercise F1.
Previous Topic
Next Session

C. LC Regression Models with repeated measurements

Introduction to Latent Class Analysis Session 2: Latent Class Regression Models C. LC Regression Models with repeated measurements

Updated Sage Article: 

          C: Repeated Measurements or Clustered Observations (pages 29-30)

LatentGOLD Tutorials:

In key driver regression models, the data file consists of a single record per respondent. Such applications often involve weak identification, meaning that it is often difficult to identify two or more latent classes. Identification is improved upon by including multiple records per respondent (repeated measures).

Three common application areas involving repeated measurements are 1) conjoint models where there are multiple ratings associated with different scenarios or product offerings (see LG Tutorial 3 and Exercise C1), 2) LC longitudinal growth models where there is a separate measurement for each point in time and the variable TIME is used as a predictor (see LG Tutorial 7A, and Exercise C2), and 3) allocation models where a certain number of points are allocated across various possible alternatives (Exercise C3).

LatentGOLD allows for the possibility of different numbers of replications Ti for each respondent case i, and provides an option to assign a different weight to each replication.

 

Exercises
Previous Topic
Back to Session
Next Session
×
×

Cart