Hypothesis Testing Using Factor Score Regression

Abstract

In this article, an overview is given of four methods to perform Factor Score Regression (FSR), namely regression FSR, Bartlett FSR, the bias avoiding method of Skrondal and Laake (2001) and the bias correcting method of Croon (2002). The bias correcting method is extended to include a reliable standard error. The four methods are compared to each other and to SEM by using analytic calculations and two Monte Carlo simulation studies to examine their finite sample characteristics. Several performance criteria are used, such as the bias using the unstandardized and standardized parameterization, efficiency, mean square error, standard error bias, type I-error rate and power. The results show that the bias correcting method, with the newly developed standard error, is the only suitable alternative for SEM. While it has a higher standard error bias than SEM, it has a comparable bias, efficiency, MSE, power and type I-error rate.

Common mechanisms of executive attention underlie executive function and effortful control in children

“…The factor score regression method results in unbiased parameter estimates in structural regression models when the factor score estimates are generated using the regression method and are used as an exogenous (i.e. independent) latent variable (Devlieger, Mayer, & Rosseel, ; Devlieger & Rosseel, ; Skrondal & Laake, ).…”

Data Integration Methods for Phenotype Harmonization in Multi-Cohort Genome-Wide Association Studies With Behavioral Outcomes

“…Certain types of factor scores used as dependent variables can lead to bias in the regression coefficients (Grice, 2001;Skrondal and Laake, 2001). Devlieger et al (2016) demonstrated that different methods of factor score calculations have similar rates of statistical power, so regression factor scores were calculated for this study.…”

Robustness of Individual Score Methods against Model Misspecification in Autoregressive Panel Models

“…Data were generated in the same way as described for simulation Study 1 except for the distribution from which the measurement residuals were drawn. In order to generate non-normal response data, measurement residuals were drawn from two different distributions as in Devlieger et al (2016): for one set of conditions, ɛ j ,tð3Þ and multiplied by the square root of the diagonal elements in Θ, resulting in curved response data that are leptokurtic (M kurtosis ¼ 9:991; SD kurtosis ¼ 14:416). For another set of conditions, ɛ j , χ 2 ð1Þ and were multiplied by the square root of the diagonal elements in Θ, resulting in positively skewed response data (M skewness ¼ 1:410; SD skewness ¼ 0:699) as may be the case when modeling responses times for instance.…”

“…Due to the development and spreading of latent variable modeling during the past decades, the focus of recent research on individual score methods has shifted away from their primary use in exploratory factor analyses towards their use in full latent variable models. In one strand of research, individual score methods are investigated with respect to their performance in multistep procedures (e. g., Croon, 2002;Devlieger, Mayer, & Rosseel, 2016;Devlieger & Rosseel, 2017;Hoshino & Bentler, 2013;Skrondal & Laake, 2001). Those approaches have in common that individual scores are first obtained based on a measurement model before they are used to study structural relationships between latent variables.…”