PERMANOVA, ANOSIM, and the Mantel test in the face of heterogeneous dispersions: What null hypothesis are you testing?

Ecological Monographs - Tập 83 Số 4 - Trang 557-574 - 2013
Marti J. Anderson1, Daniel C. I. Walsh2
1New Zealand Institute for Advanced Study (NZIAS), Massey University, Albany Campus, Private Bag 102 904, Auckland 0745 New Zealand
2Institute of Natural and Mathematical Sciences (INMS), Massey University, Albany Campus, Private Bag 102 904, Auckland 0745 New Zealand

Tóm tắt

ANOSIM, PERMANOVA, and the Mantel test are all resemblance‐based permutation methods widely used in ecology. Here, we report the results of the first simulation study, to our knowledge, specifically designed to examine the effects of heterogeneity of multivariate dispersions on the rejection rates of these tests and on a classical MANOVA test (Pillai's trace). Increasing differences in dispersion among groups were simulated under scenarios of changing sample sizes, correlation structures, error distributions, numbers of variables, and numbers of groups for balanced and unbalanced one‐way designs. The power of these tests to detect environmental impacts or natural large‐scale biogeographic gradients was also compared empirically under simulations based on parameters derived from real ecological data sets.Overall, ANOSIM and the Mantel test were very sensitive to heterogeneity in dispersions, with ANOSIM generally being more sensitive than the Mantel test. In contrast, PERMANOVA and Pillai's trace were largely unaffected by heterogeneity for balanced designs. PERMANOVA was also unaffected by differences in correlation structure, unlike Pillai's trace. For unbalanced designs, however, all of the tests were (1) too liberal when the smaller group had greater dispersion and (2) overly conservative when the larger group had greater dispersion, especially ANOSIM and the Mantel test. For simulations based on real ecological data sets, PERMANOVA was generally, but not always, more powerful than the others to detect changes in community structure, and the Mantel test was usually more powerful than ANOSIM. Both the error distributions and the resemblance measure affected results concerning power.Differences in the underlying construction of these test statistics result in important differences in the nature of the null hypothesis they are testing, their sensitivity to heterogeneity, and their power to detect important changes in ecological communities. For balanced designs, PERMANOVA and PERMDISP can be used to rigorously identify location vs. dispersion effects, respectively, in the space of the chosen resemblance measure. ANOSIM and the Mantel test can be used as more “omnibus” tests, being sensitive to differences in location, dispersion or correlation structure among groups. Unfortunately, none of the tests (PERMANOVA, Mantel, or ANOSIM) behaved reliably for unbalanced designs in the face of heterogeneity.

Từ khóa


Tài liệu tham khảo

10.1111/j.1442-9993.2001.01070.pp.x

10.1111/j.1541-0420.2005.00440.x

10.1111/j.1461-0248.2010.01552.x

10.1111/j.1461-0248.2006.00926.x

Anderson M. J, 2008, PERMANOVA+ for PRIMER: Guide to software and statistical methods

10.1111/1467-842X.00285

10.1890/0012-9658(2003)084[0511:CAOPCA]2.0.CO;2

10.1111/j.2044-8317.1987.tb00865.x

10.1093/biomet/40.3-4.318

10.1214/aoms/1177728786

10.3354/meps180257

10.1111/j.1442-9993.1993.tb00438.x

Clarke K. R, 2006, PRIMER v6: User manual/tutorial

10.3354/meps046213

10.1016/j.jembe.2005.12.017

10.1111/j.1469-1809.1941.tb02271.x

10.2307/3001535

10.2307/3001601

Edgington E. S, 1995, Randomization tests. Third edition

10.1046/j.1365-2656.2002.00606.x

10.1177/001316447403400406

Fisher R. A, 1925, Statistical methods for research workers

10.1111/j.1469-1809.1939.tb02205.x

10.3102/00346543042003237

10.1080/00949658208810568

10.1111/1467-9876.00168

10.3354/meps066285

Hayes A. F, 1996, Permutation test is not distribution free, Psychological Methods, 1, 184, 10.1037/1082-989X.1.2.184

10.1093/biomet/40.1-2.128

10.1890/0012-9615(1999)069[0001:DBRATM]2.0.CO;2

10.1111/j.1755-0998.2010.02866.x

10.1007/s004420100716

Legendre P, 1998, Numerical ecology. Second English edition

10.1080/00949650213745

10.1111/j.1095-8312.1998.tb01520.x

10.2307/2529108

Mardia K. V, 1979, Multivariate analysis

10.1890/0012-9658(2001)082[0290:FMMTCD]2.0.CO;2

10.1139/f04-051

10.1093/biomet/68.3.720

10.1080/01621459.1974.10480224

10.1037/0033-2909.86.6.1350

10.1093/biomet/23.1-2.114

10.1214/aoms/1177728599

10.2307/3236308

R Development Core Team, 2012, R: a language and environment for statistical computing

Rencher A. C, 1998, Multivariate statistical inference and applications

Romano J. P, 1990, On the behavior of randomization tests without group invariance assumption, Journal of the American Statistical Association, 85, 686, 10.1080/01621459.1990.10474928

10.1002/9780470316641

10.1016/0043-1354(90)90236-Y

10.1037/13308-000

10.1037/0033-2909.86.2.355

10.2307/1938672

10.1037/0033-2909.99.1.90

10.1080/00949650902984430

Underwood A. J, 1981, Techniques of analysis of variance in experimental marine biology and ecology, Oceanography and Marine Biology: An Annual Review, 19, 513

Underwood A. J, 1997, Experiments in ecology: their logical design and interpretation using analysis of variance

10.1007/BF00142333

10.1111/j.1541-0420.2010.01438.x

10.1890/02-0419

10.1111/j.2041-210X.2011.00127.x

10.1093/biomet/29.3-4.350

Thông tin
Thông tin xuất bản

Ecological Monographs

Tập 83 Số 4

557-574

Thông tin tác giả