PERMANOVA
PERMANOVA is a permutation test that partitions the sum of squared dissimilarities among observations into between-group and within-group parts and compares their ratio, a pseudo-F, against row permutations.
PERMANOVA는 관측값 사이 비유사도 제곱합을 집단 간과 집단 내 부분으로 나누고, 그 비율인 유사 F를 행 순열 분포와 비교하는 순열 검정입니다.
Does community composition differ between my treatment groups when the data are not multivariate normal?
데이터가 다변량 정규분포가 아닐 때 처리 집단 사이에 군집 구성이 다른지 어떻게 검정하나요?
PERMANOVA compares groups on many variables at once using any dissimilarity, such as Bray-Curtis for species counts, and reports a pseudo-F, the share of variation explained by groups (R²) and a permutation p-value. Because it is also sensitive to differences in spread, a significant result should be checked with a test of multivariate dispersion.
When to use PERMANOVA
- Comparing community or multivariate profiles among two or more groups
- Data are counts, proportions or otherwise far from multivariate normal
- A non-Euclidean dissimilarity such as Bray-Curtis or Jaccard is appropriate
When to use something else
- Responses are approximately multivariate normal with few variables: use MANOVA
- Groups clearly differ in dispersion and unbalanced sizes: interpret alongside PERMDISP or use a dispersion-robust approach
Assumptions and requirements
- Observations are exchangeable under the null hypothesis (optionally within strata)
- Similar multivariate dispersion among groups for a pure location interpretation
- The chosen dissimilarity is meaningful for the data
What the analysis reports
- Pseudo-F, R² and permutation p-value with a sums-of-squares table
- Pairwise group comparisons with Benjamini-Hochberg adjusted p-values
- A parametric dispersion check (distances to centroids)
- Bar chart of variation explained
How to report it
A PERMANOVA on Bray-Curtis dissimilarities (N = [n], [k] groups, [value] permutations) showed that group explained [value]% of the variation, pseudo-F([df1], [df2]) = [value], R² = [value], p = [value].
Related methods
Run PERMANOVA on Quorant
- Data you need
- a table (CSV or Excel) with item columns (e.g. questionnaire items) and a grouping variable
- 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.