Off-campus UNL users: To download campus access dissertations, please use the following link to log into our proxy server with your NU ID and password. When you are done browsing please remember to return to this page and log out.
Non-UNL users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Informative Hypotheses as a Methodological Middle-Ground to Detect Interpretational Confounding in Formative Measurement Models
Social scientists often seek to measure constructs that are not directly observable and are subject to measurement error. Thus, social scientists must account for measurement error while considering the relationship between the observed indicators and the construct of interest. The predominant measurement perspective views latent variables as the underlying reason that a set of indicators are related to one another. However, formative measurement provides another perspective where the indicators are thought to form or determine the construct. This measurement perspective has been heavily criticized due to several perceived limitations (e.g., interpretational confounding). Prior research has shown that interpretational confounding is not inherent to formative measurement, but may be difficult to detect using typical structural equation modeling practices. The primary objective of this study is to assess the degree to which Bayesian informed (i.e., rank ordering) hypothesis testing can detect interpretational confounding. The factors manipulated to achieve the primary objective were those that determine the possibility of interpretational confounding (i.e., the magnitude of the data generating causal-formative indicator coefficients and the informed hypothesis that is being tested) as well as those factors that may affect detection and are likely to vary in practice (i.e., multicollinearity, the number of effect indicators, and sample size). The results revealed that the Bayesian informed hypotheses could detect interpretational confounding (i.e., the mismatch between the hypothesis tested and the data generating coefficients) under the conditions considered in this study. However, higher levels of multicollinearity and smaller sample sizes did hamper this ability. Thus, in the data generating conditions considered in this study, applied researchers could use Bayesian informed hypotheses to detect interpretational confounding. Future research could conduct a sensitivity analysis to determine the lower bound of detectable differences between the causal-formative indicator coefficient magnitudes.
Lester, Houston F, "Informative Hypotheses as a Methodological Middle-Ground to Detect Interpretational Confounding in Formative Measurement Models" (2017). ETD collection for University of Nebraska-Lincoln. AAI10608702.