Partial Least Squares as applied to models with latent variables, measured indirectly by indicators, is well-known to be inconsistent. The linear compounds of indicators that PLS substitutes for the latent variables do not obey the equations that the latter satisfy. We propose simple, non-iterative corrections leading to consistent and asymptotically normal (CAN)-estimators for the loadings and for the correlations between the latent variables. Moreover, we show how to obtain CAN-estimators for the parameters of structural recursive systems of equations, containing linear and interaction terms, without the need to specify a particular joint distribution. If quadratic and higher order terms are included, the approach will produce CAN-estimators as well when predictor variables and error terms are jointly normal. We compare the adjusted PLS, denoted by PLSc, with Latent Moderated Structural Equations (LMS), using Monte Carlo studies and an empirical application.
Challenges in evaluating nonlinear effects in multiple regression analyses include reliability, validity, multicollinearity, and dichotomization of continuous variables. While reliability and validity issues are solved by employing nonlinear structural equation modeling, multicollinearity remains a problem which may even be aggravated when using latent variable approaches. Further challenges of nonlinear latent analyses comprise the distribution of latent product terms, a problem especially relevant for approaches using maximum likelihood estimation methods based on multivariate normally distributed variables, and unbiased estimates of nonlinear effects under multicollinearity. The only methods that explicitly take the nonnormality of nonlinear latent models into account are latent moderated structural equations (LMS) and quasi-maximum likelihood (QML). In a small simulation study both methods yielded unbiased parameter estimates and correct estimates of standard errors for inferential statistics. The advantages and limitations of nonlinear structural equation modeling are discussed.
1 January 2011 in European Journal of Psychological Assessment
The 15 items of the Mindful Attention and Awareness Scale (MAAS; Brown & Ryan, 2003 ) are negatively worded and assumed to assess mindfulness. However, there are indications of differences between the original MAAS and a version with the positively rephrased MAAS items (“mirror items”). The present study examines whether the mindfulness facet “mindful attention and awareness” (MAA) can be measured with both positively and negatively worded items if we take method effects due to item wording into account. To this end, the 15 negatively worded items of the MAAS and additionally 13 positively rephrased items were assessed (N = 602). Confirmatory factor analyses (CFA) models with and without regard to method effects were carried out and evaluated by means of model fit. As a result, the positively and negatively worded items should be seen as different methods that influence the construct validity of mindfulness. Furthermore, a modified version of the MAAS (MAAS-Short) with five negatively worded items (taken from the MAAS) and five positively worded items (“mirror items”) was introduced as an alternative to assess MAA. The MAAS-Short appears superior to the original MAAS. The results and the limitations of the present study are discussed.
Construct validity of the German Test Anxiety Inventory (TAI-G) was tested in two respects. Firstly, the purported four-dimensional structure of the TAI-G (comprising subscales Emotionality, Worry, Interference, and Lack of Confidence) as well as relations of the test anxiety dimensions to self-efficacy were tested. Secondly, the trait conception of the TAI-G was tested within the framework of Latent State-Trait theory. The TAI-G was given to a student sample (N=302) on three occasions with a time interval of 2 weeks along with a study-specific self-efficacy scale on occasion 1. Dimensionality assumptions as well as relations with self-efficacy were tested using cross-sectional second-order confirmatory factor analysis. The trait conception was tested separately for TAI-G subscales by specifying longitudinal confirmatory factor models (Latent State-Trait models) and by calculating variance proportions of manifest variables (Latent State-Trait coefficients) referring to different sources of systematic variance (person, situation, and method) based on parameter estimates of the models. Results were supportive of both the purported four-dimensional structure and hypothesized relationships to self-efficacy (i.e., acceptable model fit) as well as of the trait conception of test anxiety (i.e., acceptable model fit and high proportion of variance due to person component). Implications for further validation studies were discussed.