Multivariate Behavioral Research最新文献

Wording effects, the systematic method variance arising from the inconsistent responding to positively and negatively worded items of the same construct, are pervasive in the behavioral and health sciences. Although several factor modeling strategies have been proposed to mitigate their adverse effects, there is limited systematic research assessing their performance with exploratory structural equation models (ESEM). The present study evaluated the impact of different types of response bias related to wording effects (random and straight-line carelessness, acquiescence, item difficulty, and mixed) on ESEM models incorporating two popular method modeling strategies, the correlated traits-correlated methods minus one (CTC[M-1]) model and random intercept item factor analysis (RIIFA), as well as the "do nothing" approach. Five variables were manipulated using Monte Carlo methods: the type and magnitude of response bias, factor loadings, factor correlations, and sample size. Overall, the results showed that ignoring wording effects leads to poor model fit and serious distortions of the ESEM estimates. The RIIFA approach generally performed best at countering these adverse impacts and recovering unbiased factor structures, whereas the CTC(M-1) models struggled when biases affected both positively and negatively worded items. Our findings also indicated that method factors can sometimes reflect or absorb substantive variance, which may blur their associations with external variables and complicate their interpretation when embedded in broader structural models. A straightforward guide is offered to applied researchers who wish to use ESEM with mixed-worded scales.

Many psychological phenomena can be understood as arising from systems of causally connected components that evolve over time within an individual. In current empirical practice, researchers frequently study these systems by fitting statistical models to data collected at a single moment in time, that is, cross-sectional data. This raises a central question: Can cross-sectional data analysis ever yield causal insights into systems that evolve over time-and if so, under what conditions? In this paper, we address this question by introducing Equilibrium Causal Models (ECMs) to the psychological literature. ECMs are causal abstractions of an underlying dynamical system that allow for inferences about the long-term effects of interventions, permit cyclic causal relations, and can in principle be estimated from cross-sectional data, as long as information about the resting state of the system is captured by those measurements. We explain the conditions under which ECM estimation is possible, show that they allow researchers to learn about within-person processes from cross-sectional data, and discuss how tools from both the psychological measurement modeling and the causal discovery literature can inform the ways in which researchers collect and analyze their data.

The assumption of a normal distribution for latent traits is a common practice in item response theory (IRT) models. Numerous studies have demonstrated that this assumption is often inadequate, impacting the accuracy of statistical inferences in IRT models. To mitigate this issue, Gaussian mixture modeling (GMM) for latent traits, known as GMM-IRT, has been proposed. Moreover, the GMM-IRT models can also serve as powerful tools for exploring the heterogeneity of latent traits. However, the computation of GMM-IRT model estimation encounters several challenges, impeding its widespread application. The purpose of this paper is to propose a reliable and robust computing method for GMM-IRT model estimation. Specifically, we develop a mixed stochastic approximation EM (MSAEM) algorithm for estimating the three-parameter normal ogive model with GMM for latent traits (GMM-3PNO). Crucially, the GMM-3PNO is augmented to be a complete data model within the exponential family, thereby substantially streamlining the computation of the MSAEM algorithm. Furthermore, the MSAEM algorithm adeptly avoid the label-switching issue, ensuring its convergence. Finally, simulation and empirical studies are conducted to validate the performance of the MSAEM algorithm and demonstrate the superiority of the GMM-IRT models.

The reciprocal relations between sleep and affect have been a common focus in psychological research. Researchers studying affective processes often collect data multiple times a day over several days. Subjective sleep quality, on the other hand, is generally measured once at the beginning of the day. This difference in measurement frequency creates a challenge when analyzing these data, because standard dynamic models are not equipped for this. Furthermore, many of the popular approaches are based on the assumption of stationarity, meaning that processes are assumed to continue throughout the night in the same way as throughout the day. In this paper, we introduce a dynamic structural equation model that incorporates reciprocal relations between momentary affect and daily measures of sleep, tackling both of these challenges and also incorporating individual differences in these relations. To demonstrate the practical applicability of this model, we make use of an empirical example of positive and negative affect. Furthermore, we aim to give researchers the means to adapt or build on this model to align it with different research questions and other asynchronously measured variables.

Predicting ordinal responses such as school grades or rating scale data is a common task in the social and life sciences. Currently, two major streams of methodology exist for ordinal prediction: traditional statistical models such as the proportional odds model and machine learning (ML) methods such as random forest (RF) adapted to ordinal prediction. While methods from the latter stream have displayed high predictive performance, particularly for data characterized by non-linear effects, most of these methods do not support hierarchical data. As such data structures frequently occur in the social and life sciences, e.g., students nested in classes or individual measurements nested within the same person, accounting for hierarchical data is of importance for prediction in these fields. A recently proposed ML method for ordinal prediction displaying promising results for nonhierarchical data is Frequency-Adjusted Borders Ordinal Forest (fabOF). Building on an iterative expectation-maximization-type estimation procedure, I extend fabOF to hierarchical data settings in this work by proposing Mixed-Effects Frequency-Adjusted Borders Ordinal Forest (mixfabOF). The proposed method is shown to achieve performance advantages over fabOF and other existing RF-based prediction methods in settings with high random effect variability. For other settings, mixfabOF performs similarly to fabOF and alternative RF-based prediction methods.

Second-order latent growth models (LGMs) have garnered considerable attention and are increasingly utilized in longitudinal data analyses of latent constructs comprised of multiple items. The growth parameter estimates in these models are intrinsically linked to the model identification methods. Latent-standardization (identification) methods, in which the latent variable is standardized at a reference time point (e.g., eta-1), yield theoretically unique and interpretable growth parameters. Traditional latent-standardization methods indirectly standardize eta-1 via the first-order component of the second-order LGM by constraining item intercepts and/or loadings. Such methods require a two-step modeling procedure and do not truly standardize eta-1. This article proposes a 1-stage method that indirectly standardizes eta-1 through the second-order component of the model by constraining the mean and variance of the level factor. This new single-step modeling method ensures eta-1 is truly standardized, with a mean of 0 and a variance of 1. Theoretical, simulated, and empirical comparisons are conducted across different latent-standardization methods, demonstrating the target accuracy and implementation simplicity of the proposed 1-stage method.

Many applications of network modeling involve cross-sectional data of psychological variables (e.g., symptoms for psychological disorders), and analyses are often conducted using a regularized Gaussian graphical model (GGM) employing a lasso, also known as the graphical lasso or glasso. Appropriate methodology for handling missing data is underdeveloped while using glasso, precluding the use of planned missing data designs to reduce participant fatigue. In this research, we compare three approaches to handling missing data with glasso. The first resembles a two-stage estimation approach-borrowed from the covariance structure modeling literature-whereby a saturated covariance matrix among the items is estimated prior to using glasso. The second and third approaches use glasso and the expectation-maximization (EM) algorithm in a single stage and either use EBIC or cross-validation for tuning parameter selection. We compared these approaches in a simulation study with a variety of sample sizes, proportions of missing data, and network saturation. An example with data from the Patient Reported Outcomes Measurement Information System is also provided. The EM algorithm with cross-validation performed best, but all methods appeared to be viable strategies under larger samples and with less missing data.

Longitudinal designs afford the opportunity to examine the many different ways in which variables can be related over time, which can be both a blessing and a curse. Much has been written about the need to distinguish between-person relations of individual mean differences from within-person relations of time-specific residuals for time-varying predictors. The present work expands on this topic by describing the need to further distinguish between-person relations among individual slopes for change over time. Using simulation methods, this problem is demonstrated within univariate longitudinal models (i.e., multilevel or mixed-effects models using observed predictors), as well as in multivariate longitudinal models (i.e., structural equation models using latent predictors). The discussion presents recommendations for practice, along with caveats and concerns regarding related longitudinal models for lead-lag effects.

Interpersonal synchronization is a concept often studied in psychology. Whereas most research focuses on dyads, triadic systems such as family triads warrant increased attention. A crucial challenge in taking a triadic view on synchronization is how to quantify it, since a statistical measure that captures the level of triadic synchronization in one value, while discarding dyadic synchronization only, is lacking so far. The current paper therefore investigated three existing measures that show potential to capture triadic synchronization and proposes two novel ones. We also present a significance test that allows to investigate whether the observed triadic synchronization in a triad is stronger than can be expected by chance, while accounting for potential auto-dependence in the data. By means of a simulation study, we tested (1) how the measures react to different potential synchronization patterns; (2) the Type I error rate and the power of the significance test. The results showed that only one measure, i.e., the newly proposed adapted multiplication of pairwise correlations (AMPC), can effectively capture triadic synchronization, while discarding dyadic synchronization. We then applied the AMPC measure to intensive longitudinal data on attachment-related measures in families, showing that AMPC can detect meaningful triadic synchronization in empirical data.

This study primarily investigates the impact of ignoring nonnormal distributions in RSEM models on the estimation of parameters in the second residual structure. The results of the simulation studies demonstrate that when the RSEM model follows a nonnormal distribution, it is crucial to test and estimate the nonnormal distribution while constructing mixture RI-AR or mixture RI-CLPM models. This approach guarantees the unbiased estimation of autoregressive parameters and cross-lagged parameters in the second residual structure. If, during the construction of an empirical model, the nonnormal distribution of mixture RI-AR models or mixture RI-CLPM models is not taken into account, or if a normal distribution is assumed directly for analysis, the resulting parameter estimates for autoregressive parameters and cross-lagged parameters will be biased, leading to erroneous inferences.