Mantel and Partial Mantel Tests

The Mantel test correlates the corresponding elements of two distance matrices and judges significance by permuting the rows and columns of one matrix; the partial Mantel test controls for a third matrix.

Mantel 검정은 두 거리행렬의 대응 원소 사이 상관을 구하고, 한 행렬의 행과 열을 순열하여 유의성을 판단하며, 부분 Mantel 검정은 세 번째 행렬을 통제합니다.

Are sites that differ more in environment also more different in species composition?

환경이 더 다른 지점일수록 종 구성도 더 다른가요?

The Mantel test computes a Pearson or Spearman correlation between all pairwise dissimilarities of one data set and all pairwise distances of another, with a permutation p-value that respects the dependence among pairs. A partial Mantel test removes the effect of a third distance matrix, and a Mantel correlogram shows how similarity changes across distance classes.

When to use Mantel and Partial Mantel Tests

• Your hypothesis is stated in terms of distances, such as community dissimilarity versus environmental or geographic distance
• You want to control for a third distance matrix
• You want a correlogram of similarity across distance classes

When to use something else

• The hypothesis concerns the raw data rather than distances: use Redundancy Analysis (RDA), which has more power
• You compare two configurations of the same objects: use Procrustes analysis

Assumptions and requirements

• Linear (Pearson) or monotonic (Spearman) relation between distances
• Objects are exchangeable under the null hypothesis
• Spatial autocorrelation can inflate partial Mantel type I error

What the analysis reports

• Mantel r and permutation p-value
• Partial Mantel r and p-value when control variables are given
• Mantel correlogram with Holm-adjusted p-values

How to report it

A Mantel test (Pearson, [value] permutations) between community dissimilarity and environmental distance (N = [n]) gave r_M = [value], p = [value].

Related methods

• Redundancy Analysis (RDA)
• Procrustes Analysis
• Moran's I

Run Mantel and Partial Mantel Tests on Quorant