Recovery of predictor relationships via semiparametric and parametric growth models under misspecification Public Deposited
Downloadable ContentDownload PDF
- Last Modified
- March 22, 2019
Sterba, Sonya Kourany
- Affiliation: College of Arts and Sciences, Department of Psychology and Neuroscience
- Proper specification of the discrete vs. continuous nature of individual differences in growth is notoriously error prone. Despite this fact, psychologists desire accurate prediction of growth trajectories. In this light, the present study asks: can parametric or semiparametric methods better recover relationships between predictors and individual trajectories when the nature of individual differences is misspecified? In the first part of this study, novel approaches were adopted to ensure that neither the measure of predicted relationship recovery nor the choice of model specification favored a particular combination of generating model (discrete vs. continuous) and fitted model (parametric vs. semiparametric). Results indicated that the increase in mean squared error (MSE) associated with misspecifying classes as continua was greater than the increase in MSE associated with misspecifying continua as classes. In fact, a misspecified parametric model had better MSE than a correctly specified semiparametric model at low N, due to the former's large efficiency advantage and the latter's small advantage in bias. In the second part of this study, I investigated whether the preferability of a parametric vs. semiparametric model could be altered by invoking generating conditions commonly thought to be unfavorable to a particular fitted model. The parametric model continued to have better predicted relationship recovery under both unfavorable conditions, contradicting conventional wisdom. Discussion focuses on the implication of these results for practice.
- Date of publication
- May 2010
- Resource type
- Rights statement
- In Copyright
- "... in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Psychology (Quantitative)."
- Bauer, Daniel
- Degree granting institution
- University of North Carolina at Chapel Hill
- Place of publication
- Chapel Hill, NC
- Open access
This work has no parents.
|Recovery of predictor relationships via semiparametric and parametric growth models under misspecification||2019-04-11||Public||