Marine Biodiversity and Ecosystem Functioning
EU Network of Excellence

 
Main Menu

· Home
· Contacts
· Data System
· Documents
· Events Calendar
· FAQ
· Forums
· Job M@RKET
· Links
· MarBEF Open Archive
· Network Description
· Outreach
· Photo Gallery
· Quality Assurance
· Register of Resources
· Research Projects
· Rules and Guidelines
· Training
· Weekly News Bulletin
· Wiki
· Worldconference
 

 

Last 3 Job opportunities View latest job opportunities RSS feed


>> Go to Job M@RKET

Last 2 forum posts View MarBEF Forum RSS feed

 Field course on sharks in South Africa (June 2011)
 training opportunity in basic taxonomy

>> Go to forums

 

 
Who's Online

Currently 43 guest(s) online
Currently 0 member(s) online
 

 
Login

Username

Password

 

Event Calendar

· Year View
· Month View
· Add an Event
· Add a Deadline
Jump to the following view:
 

Detecting changes in biodiversity: experimental design and data analysis



Location: University of Pisa
Event Type: Training Course
Event Date: 2006-09-04    
End Date: 2006-09-15

Event Description:
Complex experimental designs are crucial for most studies of natural systems. The course covers the fundamentals in the design and analysis of experiments in ecology, with emphasis on methods to assess changes in biodiversity. The participants will be instructed on how to decide what kind of analysis is most appropriate given the hypothesis of interest and how to interpret their results. The course will consist of lectures and practical PC exercises. Applications must include a CV.

This course is endorsed by MarBEF under the general conditions, i.e. per MarBEF member one participant is allowed to book the costs of Travel and Subsistence up to a maximum of 750 Euro on their MarBEF account.



Contact email: bencecc@discat.unipi.it

Event programme:
MULTIVARIATE COMMUNITY ANALYSES USING PRIMER* (K.R. Clarke, R.N. Gorley, Plymouth Marine Laboratory and PRIMER-E, UK)

Similarity, clustering and ordination
Comparative definitions of similarity/dissimilarity/distance between two assemblages, to satisfy desirable biological criteria (including coefficient and transformation choices; standardisation/normalisation choices; differential species weighting in similarity definition, e.g. from spatial clumping models fitted to replicates). Also, multivariate analysis using ‘biodiversity-based’ dissimilarity measures, i.e. exploiting taxonomic or phylogenetic relatedness of the species in two samples.
Clustering (i.e. grouping) and ordination (i.e. 'mapping') of inter-sample similarities, usually in 2 or 3-d plots (including diagnostics for adequacy of low dimensional display)
Practical sessions:
Clustering and ordination (simple hierarchical clustering and Principal Components Analysis/non-metric Multi-Dimensional Scaling, in contrast and combination; linking of species information to sample patterns)

Multivariate testing
Multivariate non-parametric permutation tests on similarity matrices ('analysis of similarity' tests, analogues of ANOVA for defined sample structure); simple power considerations). Also tests for the presence of structure in a priori unstructured samples (similarity profile tests))
Practical sessions:
Multivariate ANOSIM tests (simple 1- and 2-way crossed and nested layouts) for determining and quantifying differences between a priori defined groups of samples, or unstructured data (SIMPROF tests). Species responsible for identified groupings.

Contrast with univariate analyses
Univariate assemblage summaries (diversity measures, their sampling properties and information content in relation to multispecies patterns);
Graphical summaries (e.g. dominance plots, including testing for defined sample structure);
Biodiversity measures based on taxonomic/phylogenetic distinctness in a sample (including sampling properties and simulation testing framework).

Practical sessions:
Standard diversity indices, dominance and ABC plots (description and testing, including multivariate display);
Univariate taxonomic distinctness measures (including tests against expectation from master taxon lists)

Comparing multivariate patterns
Linking (multivariate analysis of) environmental variables to biotic assemblage patterns (e.g. by searching over all combinations of abiotic variables, the BEST procedure, including a global test for significance of the resulting relation). Also, missing data estimation for environmental matrices
Generalising this comparison procedure (stepwise version of search for matching biotic-biotic, environmental-biotic, model-biotic matrices, etc.)
Testing multivariate biotic patterns against particular models (e.g. time/space trends or seasonal cycles) using rank permutation tests (Mantel type);
Summarising multiple multivariate analyses, e.g. many different ordinations, with 2nd stage analyses (visualising relative effects of transformation choice, taxonomic identification level and chosen similarity measure on the multivariate pattern; describing and testing for changes in internal structure of ordinations, in some repeated measures designs for example).
Practical sessions:
Multivariate analysis of environmental variables, and choosing a subset of variables which 'best explains' the biotic assemblage pattern;
Identifying subsets of species (e.g. surrogates) which between them capture the sample pattern of the full assemblage;
Testing for significant serial change in an assemblage along a gradient in space (or time);
Displaying relationships between multivariate patterns from different analysis choices, and examining differences in (assemblage) spatial patterns over a time course

*PRIMER: Plymouth Routines In Multivariate Ecological Research (see website www.primer-e.com)


A GENERAL INTRODUCTION TO EXPERIMENTAL DESIGN (L. Benedetti-Cecchi), University of Pisa, Italy)

Logical and philosophical frameworks for the analysis of ecological complexity
Falsification and the hypothetic-deductive approach; Strong inference; Bayesian inference

Sampling populations
Ecological variables; Frequency distributions; Parameters and their estimates; Precision and accuracy of sample estimates

Relationships among ecological models, statistical models and data
Estimation and hypothesis testing; Linear statistical models; Methods of estimation: OLS and ML; Statistical hypothesis testing: general hints.
Practical sessions:
Sampling populations: influence of variance and sample size on sample estimates.

Experimental design
Basic concepts: replication, randomisation, independence; Choosing levels for predictor variables: fixed vs. random factors; Relationships among predictor variables: hierarchical and factorial designs; Extension to multifactorial designs. Expected Mean Squares in mixed-models designs.
Practical sessions:
Assessing spatial and temporal variability in biodiversity: the analysis of pattern using simulated data sets.

Experimental design
Hierarchical designs: solution to spatial and temporal confounding; Hierarchical designs: sampling at multiple scales in space and time; Factorial designs: understanding interactions among predictor variables; Assessing the generality (or lack thereof) of ecological processes: multifactorial experiments. The experimental analysis of biodiversity: designs that can separate sampling and complementarity effects while controlling for changes in abundance of species.
Practical sessions:
Understanding ecological processes: analysis and interpretation of real ecological experiments. Effects of biodiversity on temporal stability of marine assemblages.


ANALYSING MULTI-SPECIES RESPONSES TO COMPLEX EXPERIMENTAL DESIGNS (M. J. Anderson, University of Auckland, New Zealand)

Permutational multivariate analysis of variance (PERMANOVA) and tests of homogeneity of multivariate dispersions.
Solution based on inter-point dissimilarities; Interpreting multivariate interaction terms; Comparison with other methods, such as ANOSIM, including assumptions; Additional notes on the nature of multivariate data and dissimilarity measures; The effect of choice of dissimilarity measure and transformations on relative dispersions.
Practical sessions (M.J. Anderson and R. N. Gorley):
PERMANOVA and HOMOGENEITY: Analysing two-way models with interaction, including a posteriori contrasts; Analysing two-factor nested designs; Distinguishing effects on dispersions and effects on locations in multivariate space; Principal coordinate analysis (PCO), as an unconstrained ordination method for viewing patterns.

Analysing multivariate responses to complex experimental designs.
Extending two-factor PERMANOVA to more complex experimental designs, including fixed and random factors as well as crossed and nested structures in mixed models; Appropriate permutation procedures for complex designs by reference to the expected mean squares (EMS) and construction of F-ratios; Estimating pseudo multivariate variance components; Choosing the correct units to permute; Monte Carlo approximation when there are not enough permutations for the test.
Practical sessions (M.J. Anderson and R. N. Gorley):
PERMANOVA for complex experimental designs; Constructing the correct design file for input; Logical procedures once the full partitioning has been done; A posteriori comparisons and appropriate plots for unravelling and interpreting effects obtained from multi-factorial designs.

Multivariate multiple regression based on dissimilarities.
Relating multivariate response data to quantitative predictors (e.g., environmental variables). Analysis of a dissimilarity matrix on the basis of any linear model; Redundancy analysis and distance-based redundancy analysis; Model selection procedures and multivariate analogues to the univariate AIC and BIC criteria.
Practical sessions (M.J. Anderson and R. N. Gorley):
DISTLM and sequential model fitting; forward selection, backwards elimination and step-wise procedures; Redundancy analysis and associated plots from a DISTLM.

Canonical analysis of principal coordinates (CAP), a constrained ordination method.
Unconstrained versus constrained ordination techniques - distinctions and potential pitfalls;
Canonical discriminant analysis (CDA) based on dissimlarities; Cross-validation check on arbitrariness of results; Comparison with other methods, including PERMANOVA; Canonical correlation analysis (CCorA) based on dissimilarities.
Practical sessions (M.J. Anderson and R. N. Gorley):
CAP; Finding axes in the multivariate space that are best at discriminating among a priori groups; Assessing the distinctness of groups using cross-validation; Relatinoships between canonical axes and the original (e.g. species) variables; Finding axes in the multivariate space having the strongest correlation with environmental variables; Using canonical axes to place new observations along an environmental gradient, based on species data.


Putting it all together - a general strategy for multivariate analysis.
Partitioning according to the full model; Tests of hypotheses; Unconstrained and constrained ordination; Identifying original species which are driving the patterns.
Practical sessions:
Beginning only with the experimental design and the data, complete an entire multivariate analysis in response to a complex experimental design, identify important factors and interactions, display appropriate plots, determine which species may be driving patterns.

Environmental impact assessment
Environmental analysis and the precautionary principle; Building models and analysing multivariate data from BACI and beyond-BACI experimental designs; Environmental monitoring and multivariate control charts based on dissimilarities; Bootstrap results.
Practical sessions (M.J. Anderson and R. N. Gorley):
Build a complete multivariate beyond-BACI model, using PERMANOVA and DISTLM; Use the CONTROLCHART computer program to monitor multivariate ecosystems and highlight when a stable system is going “out of control”.

Collaborators and organisers:
Contact address: Prof. Lisandro Benedetti-Cecchi, Department of Biology, Via A. Volta 6, I-56126, Pisa, Italy. Office: +39 50 2219013; Fax: +39 50 49694; e-mail: bencecc@discat.unipi.it

Website url: 

Registration form: 

Registration fee: 1500,00

Relevant costs that should be taken by MARBEF:  

Possibility for day-care centres (facilities for children): No

MarBEF supported event: Yes

Event within Framework of MARBEF: (none)

Maximum number of participants: 30

External participants (non MARBEF) allowed: Yes

Posted by Benedetti-Cecchi_Lisandro on 2006-04-10 and approved by webmaster