Profile Analysis (Repeated-Measures MANOVA)

Profile analysis is a multivariate approach to repeated measures that tests parallelism, levels and flatness of group profiles using differences between adjacent occasions, without assuming sphericity.

프로파일 분석은 인접 측정 시점 간 차이를 이용해 집단 프로파일의 평행성, 수준, 평탄성을 구형성 가정 없이 검정하는 반복측정의 다변량 접근법입니다.

Do my groups change in the same way across repeated measurements, and do they differ in overall level?

반복 측정에서 집단들이 같은 모양으로 변하는지, 전체 수준이 다른지 알고 싶어요.

Profile analysis converts repeated measures into differences between adjacent occasions and tests whether group profiles are parallel with a MANOVA on those differences. It then tests whether groups differ in overall level with an ANOVA on each person's mean and whether the combined profile is flat with a Hotelling T² test, so sphericity is not required.

When to use Profile Analysis (Repeated-Measures MANOVA)

• The same people are measured on several occasions or commensurate scales
• Two or more groups are compared
• Sphericity is doubtful and a multivariate approach is preferred

When to use something else

• Measures are on different scales and not commensurate: use MANOVA
• There are missing occasions or time-varying covariates: use a linear mixed model

Assumptions and requirements

• Measures share a common scale
• Multivariate normality of difference scores
• Homogeneous covariance matrices of differences across groups (Box's M is reported)
• Complete data on all occasions (incomplete rows are dropped)

What the analysis reports

• Parallelism test with four multivariate statistics
• Levels test (F) with eta squared
• Flatness test (Wilks' lambda, Hotelling T², F)
• Group means at each occasion

How to report it

Profile analysis of [p] occasions across [k] groups (N = [n]) showed that the profiles were [not] parallel, Wilks' Λ = [value], F([df1], [df2]) = [value], p = [value]; the levels test gave F([df1], [df2]) = [value], p = [value], and the flatness test gave F([df1], [df2]) = [value], p = [value].

Related methods

• Repeated Measures ANOVA
• MANOVA
• Linear Mixed Model

Run Profile Analysis (Repeated-Measures MANOVA) on Quorant