Detrended correspondence analysis - Wikipedia

statisticsecologymultivariate-analysisordination

Abstraction: Ordination method correcting arch and edge artifacts in ecological gradient data

Key points:

DCA is a multivariate method for species-rich sparse ecological community matrices, created by Mark Hill in 1979 and implemented in DECORANA (FORTRAN)
Corrects two artifacts of standard correspondence analysis: "arch effect" (second axis curves from first) and "edge effect" (variance compressed at ends)
Algorithm divides the first CA axis into segments (default 26), rescales each to zero mean on second axis, flattening the curve
Rule of thumb: 4 DCA units = total community turnover; 1 unit approximates constant turnover rate
No significance tests available; constrained version DCCA allows Monte Carlo permutation testing against a null model
Available in R's vegan package (decorana function)

Connections: Correspondence Analysis · Ordination · Dimensionality Reduction · Multivariate Analysis