UBC Theses and Dissertations
On modelling change and growth when the measures themselves change across waves : methodological and measurement issues and a novel non-parametric solution Lloyd, Jennifer Elizabeth Victoria
In the past 20 years, the analysis of individual change has become a key focus of research in education and the social sciences. There are several parametric methodologies that centre upon quantifying change. These varied methodologies, known as repeated measures analyses, are commonly used in three research scenarios: In Scenario 1, the exact same measure is used and re-used across waves (testing occasions). In Scenario 2, most of the measures' content changes across waves - typically commensurate with the age and experiences of the test-takers - but the measures retain one or more common items (test questions) across waves. In Scenario 3, the measures either vary completely across waves (i.e., there are no common items) or the sample being tested across waves is small or there is no norming group. Some researchers assert that repeated measures analyses should only occur if the measure itself remains unchanged across waves, arguing that it is not possible to link or connect the scores (either methodologically or conceptually) of measures whose content varies across waves. Because it is not uncommon to face Scenarios 2 and 3 in educational and social science research settings, however, it is vital to explore more fully the problem of analysing change and growth with measures that vary across waves. To this end, the first objective of this dissertation is to weave together the (a) test linking and (b) change/growth literatures for the purpose of exploring this problem in a comprehensive manner. The second objective is to introduce a novel solution to the problem: the nonparametric hierarchical linear model (for multi-wave data) and the non-parametric difference score (for two-wave data). Two case studies that demonstrate the application of the respective solutions are presented, accompanied by a discussion o f the novel solution's strengths and limitations. Also presented is a discussion about what is meant by 'change'.
Item Citations and Data