Evaluating the Predictive Power of Ordination Methods in Ecological Context

When striving for the ordination methods best predicting independently measured site factors, the following questions arise: does the optimal choice depend on the kind of biological community analysed? Are there different ordination methods needed to address different site factors? Simultaneously, I explore alternative similarity approaches of entire ordinations, as well as the role of the transformations applied to the scale used in measuring species performance. The combination of methods and data transformations results in 96 alternative solutions for any one data set. These are compared by a graphical display, that is, an ordination of ordinations. The goodness-of-fit of independently measured site factors is assessed by two alternative methods. The resulting 96 performance values serve as independent variables in trend surfaces overlaid to the ordination of ordinations. The results from two real-world data sets indicate that some ordination methods greatly vary with data transformation. Scores close to a binary scale perform best in almost all ordination methods. Methods that intrinsically constrain the range of species scores, such as principal components analysis based on correlation, correspondence analysis (including its detrended version), nonmetric multidimensional scaling, as well as principal coordinates analysis based on the Bray-Curtis distance, always figure among the most successful methods, irrespective of data used.

[1]  Jari Oksanen,et al.  Instability of ordination results under changes in input data order: explanations and remedies , 1997 .

[2]  László Orlóci,et al.  Multivariate Analysis in Vegetation Research , 1975 .

[3]  S. Dray,et al.  Assessing species and community functional responses to environmental gradients: which multivariate methods? , 2012 .

[4]  E. van der Maarel,et al.  Transformation of cover-abundance values in phytosociology and its effects on community similarity , 1979, Vegetatio.

[5]  K. R. Clarke,et al.  Non‐parametric multivariate analyses of changes in community structure , 1993 .

[6]  Marti J. Anderson,et al.  A new method for non-parametric multivariate analysis of variance in ecology , 2001 .

[7]  Donald A. Jackson,et al.  How well do multivariate data sets match? The advantages of a Procrustean superimposition approach over the Mantel test , 2001, Oecologia.

[8]  M. Hill,et al.  Detrended correspondence analysis: an improved ordination technique , 1980 .

[9]  Henrik von Wehrden,et al.  Pluralism and diversity: trends in the use and application of ordination methods 1990-2007. , 2009 .

[10]  M. Austin.  Vegetation and Environment: Discontinuities and Continuities , 2013 .

[11]  J. M. A. Swan,et al.  An Examination of Some Ordination Problems By Use of Simulated Vegetational Data , 1970 .

[12]  Jan Lepš,et al.  Multivariate Analysis of Ecological Data using CANOCO , 2003 .

[13]  R. Fisher.  THE PRECISION OF DISCRIMINANT FUNCTIONS , 1940 .

[14]  N. Mantel.  The detection of disease clustering and a generalized regression approach. , 1967, Cancer research.

[15]  J. Podani.  Braun-Blanquet's legacy and data analysis in vegetation science , 2006 .

[16]  Christina Gloeckner,et al.  Modern Applied Statistics With S , 2003 .

[17]  János Podani,et al.  Assessing the relative importance of methodological decisions in classifications of vegetation data , 2015 .

[18]  John C. Gower,et al.  Statistical methods of comparing different multivariate analyses of the same data , 1971 .

[19]  P. Legendre,et al.  Multivariate approach to study species assemblages at large spatiotemporal scales : the community structure of the epibenthic fauna of the estuary and Gulf of St. Lawrence , 1990 .

[20]  D. Goodall,et al.  Objective methods for the classification of vegetation. III. An essay in the use of factor analysis , 1954 .

[21]  Mike P. Austin.  Inconsistencies between theory and methodology: a recurrent problem in ordination studies , 2013 .

[22]  L. Ruokolainen,et al.  Differences in performance of four ordination methods on a complex vegetation dataset , 2006 .

[23]  P. Legendre,et al.  RELATING BEHAVIOR TO HABITAT: SOLUTIONS TO THEFOURTH-CORNER PROBLEM , 1997 .

[24]  O. Wildi.  Data Analysis in Vegetation Ecology , 2010 .

[25]  Peter R. Minchin,et al.  An evaluation of the relative robustness of techniques for ecological ordination , 1987 .

[26]  E. Feoli,et al.  Can similarity theory contribute to the development of a general theory of the plant community , 2011 .

[27]  J. Gower.  Some distance properties of latent root and vector methods used in multivariate analysis , 1966 .

[28]  Donald A. Jackson,et al.  How many principal components? stopping rules for determining the number of non-trivial axes revisited , 2005, Comput. Stat. Data Anal..