Latent Class Analysis (LCA)

Latent class analysis (LCA) is a mixture model that identifies unobserved subgroups in a population from patterns of responses on categorical indicators.

잠재계층분석(LCA)은 범주형 지표에 대한 응답 패턴으로부터 모집단 안의 관측되지 않은 하위집단을 찾아내는 혼합모형이다.

How do I find hidden subgroups of respondents based on their answers to several categorical questions?

여러 범주형 문항에 대한 응답으로 숨은 응답자 하위집단을 어떻게 찾나요?

Latent class analysis finds unobserved subgroups whose members share similar response patterns on categorical items, assigning each case a probability of belonging to each class. It fits person-centered typologies, with the number of classes chosen using BIC, entropy and interpretability.

When to use Latent Class Analysis (LCA)

• Several binary or categorical survey items
• Building a typology of respondents, patients or customers
• Person-centered analysis rather than variable-centered analysis

When to use something else

• Continuous indicators; use latent profile analysis
• Grouping on distances without a probabilistic model; use k-means or hierarchical clustering

Assumptions and requirements

• Local independence of indicators within each class
• Categorical indicators
• Adequate sample size for the number of classes
• Multiple random starts to avoid local maxima

What the analysis reports

• Fit statistics for each class solution (AIC, BIC, sample-size adjusted BIC)
• Entropy
• Class proportions
• Item-response probabilities by class
• Posterior class membership probabilities

How to report it

A [k]-class solution was selected based on the lowest BIC ([value]) and an entropy of [value].

Related methods

• Latent Profile Analysis
• K-Means Clustering
• Mixture Modeling

Run Latent Class Analysis (LCA) on Quorant