Canonical Correlation Analysis
Canonical correlation analysis finds pairs of weighted composites of two variable sets that are maximally correlated with each other, summarizing the multivariate association between the sets.
정준상관분석은 두 변수 묶음 각각의 가중 합성점수 쌍을 서로 상관이 가장 크도록 찾아 두 묶음 사이의 다변량 관련성을 요약하는 기법입니다.
How strongly is my set of attitude scores related to my set of behavior measures, and which variables drive that relationship?
태도 척도 묶음과 행동 측정치 묶음이 전체적으로 얼마나 관련되어 있고, 어떤 변수가 그 관련을 이끄는지 알고 싶어요.
Canonical correlation analysis estimates successive pairs of linear composites, one from each variable set, whose correlations are as large as possible and uncorrelated with earlier pairs. Wilks' lambda with Bartlett's chi-square tests whether the remaining canonical correlations are jointly zero, and structure loadings show which variables define each function.
When to use Canonical Correlation Analysis
- You have two sets of continuous variables and want to describe the association between the sets as a whole.
- You want to know how many independent dimensions link the two sets.
- Several outcomes and several predictors are measured on the same cases and none is singled out as the main outcome.
When to use something else
- There is only one outcome variable; use multiple linear regression instead.
- The goal is to test group differences on several outcomes; use MANOVA instead.
Assumptions and requirements
- Linear relationships within and between the variable sets.
- Multivariate normality for the significance tests.
- No perfect multicollinearity within either set.
- A sample size of roughly ten or more cases per variable.
What the analysis reports
- Canonical correlations and squared canonical correlations
- Wilks' lambda, Bartlett chi-square and p for each sequential test
- Standardized canonical coefficients
- Structure loadings for both variable sets
How to report it
A canonical correlation analysis between [p] predictor and [q] criterion variables (N = [n]) was significant, Wilks' Λ = [value], χ²([df]) = [value], p = [value], with a first canonical correlation of Rc = [value].
Related methods
Run Canonical Correlation Analysis on Quorant
- Data you need
- a table (CSV or Excel) with predictor (feature) columns and item columns (e.g. questionnaire items)
- Price
- 400 credits for one analysis up to 10,000 rows (complexity level 3); larger or wider data costs more. Optional methodologist review: 1800 credits.
- Checking
- Checked by property tests in every deployment
- Before you pay
- A free pre-analysis diagnosis checks your data and gives an exact quote; nothing is deducted until you approve it.