New permutation methodologies to deal with the multiplicity issue: multiple comparisons and multiple tests with applications to single-case experiments and to regression analysis
Year: 2012
Type: dissertation
Abstract: The thesis presents new results within the permutation testing approach in order to deal with real complex problems. Very often real datasets are the result of complicated planning phases of the study or they are complex by themselves. Multiple outcomes are often of interest and, a fact which increases further on their complexity, complicated and unknown dependence structures can underlie such multivariate responses (i.e. the multiplicity issue). Two particular applied problems are faced: single-case experiments and regression analysis of ordinal data. Both univariate and multivariate solutions to such issues are proposed in this thesis, which show to successfully handle the data complexity by means of permutation tests and their nonparametric combination. Regarding the single-case experiments problem a complex solution is developed which exploits the joint use of smoothing techniques and permutation theory. For ordinal data analysis instead, we propose some permutation solutions that use parametric estimates as test statistics, creating a link between parametric and nonparametric problem solving. Several simulation studies and real case applications show the good behavior and the usefulness of the presented procedures.
Author: Francesca Solmi
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2276655326