Instructor: Dr. Georg Roth (Ur- und Frühgeschichte, University of Cologne)
The workshop focuses on explorative and inferential multivariate methods suited specifically for archaeological data sets. Methods are discussed with special reference to archaeological research questions, i.e. what insight is sought for and which method can support this approach.
Starting with the classic archaeological method CA its fields of application are reviewed with a focus on its limitations and an alternative solution for archaeological data, transformation based PCA (tbPCA), is introduced. First day ends with including explanatory information into ordinations i.e. canonical ordinations (CCA, tbRDA).
Besides (explorative) ordination finding reliable groupings is an important field of multivariate analysis. Starting with a special emphasis on choosing suitable distance measures traditional cluster approaches are discussed and enhanced by building reliable groupings based on internal validity criteria. The workshop closes with perspectives on multivariate methods for more specific purposes.
During afternoon intermissions individual mentoring is possible. Participation requires base knowledge of data handling in R. Participants are kindly asked to refresh their skills in advance. A small list of free media is provided in advance.
10.00-12.00h Introduction: Why m-var stats anyway?
Repetition getting data into R, multivariate stats packages, CA – myths of CA and seriation.
13.00-15.00h Comparing CA with tbPCA; new alternative ordinations for archaeological data sets.
15.00-16.00h Didactic break (individual mentoring)
16.00-18.00h Testing interpretations in ordination (constrained analysis); how to connect cause and effect for multivariate data.
10.00-12.00h Distance based methods with focus on clustering; how to measure differences for different kinds of data.
13.00-15.00h Finding reliable groups with cluster validity criteria; how to define objective groups.
15.00-16.00h Didactic break (individual mentoring).
16.00-18.00h Perspectives on multivariate solutions for specific research questions.
Borcard et al. 2011: D. Borcard/Fr. Gillet/P. Legendre, Numerical ecology with R (New York 2011).
Legendre/Legendre 2012: P. Legendre/L. Legendre, Numerical ecology. Developments in environmental modelling 24 (Amsterdam 2012 3rd).
Date, Time: 03/06/2016, 10:00 h – 18:00 h
& 10/06/2016 10.00 – 18.00 h