Chained equations

Rebecca Kuiper (PhD)
Department of Methodology & Statistics
Faculty of Social Sciences
Utrecht University
P.O. Box 80140, 3508 TC Utrecht
Phone: +31 30 253 1571 / 4438 (secretary)
E-mail: Rebecca Kuiper

Supervisors: Prof. dr H.J.A. Hoijtink

Project running from: 1 May 2007 – 1 May 2012

Summary:
Theories often have multiple implications that have to be evaluated. Multiple hypotheses addressing different variables are not easily summarized in one statistical model, because often it is too complicated to account for the dependencies between the variables. Multiple hypotheses are usually evaluated separately which increases the probability of errors of the first kind and/or reduces the power. See, for example, Toothaker (1993), Benjamini and Hochberg (1995) and Maxwell (2004) for a discussion of these matters. In this project chained equations (van Buuren, Boshuizen, Knook, 1999; Raghunathan, Lepkowski, Van Hoewyk, and Solenberger, 2001; Buuren, Brand, Groothuis-Oudshoorn and Rubin, to appear) will be used to build statistical models for multiple hypotheses addressing the same or different data sets. Chained equations have thus far been used for multiple imputation of missing values. Here they will be used to build one statistical model for the evaluation of multiple hypotheses.

Date of defence:  27 January 2012

Title of thesis:  Model selection: How to evaluate order restrictions (ISBN: 978-90-393-5715-6)

Promotor:  Prof. dr H.J.A. Hoijtink