Livio Finos Conference 2012-01-12 4 pm
Professor, Padova University Statistical Science Dept.
"Multivariate permutation tests for analysis of kinematic data"
Psychophysic experiments are often characterized by many trials on few subjects. The hierarchical structure of the data (i.e. repeated measures on the same subject) make the mixed model an ideal tool to make statistical analysis. However, the small sample size (i.e. few subjects) imposes strong assumptions for the validity of classical tools; hence making the results less usefull.
Moreover, kinematic analysis supposed that many variables are measured/collected in the same trial, making the problem multivariate in nature. This case is hard to deal within the classical parametric framework and the researcher is usually forced to provide separated analysis for each variable, then loosing the multivariate perspective of the problem and encountering the problem of un corrected p value.
In this talk we present a multivariate permutation approach for the analysis of kinematic data which requires less assumptions than classical methods and allow for multivariate (i.e. global) and univariate (i.e. in detail) inference within the same methodological framework. An application to real data is shown and discussed.