Multivariate Behavioral Research最新文献

Following Kelderman and Molenaar's demonstration that a factor model with person specific factor loadings is almost indistinguishable from the standard factor model in terms of overall fit, we examined person specific measurement models in Item Response Theory, person specific discrimination and difficulty parameters were created by adding random variation at the item by person level. Using standard fitting algorithms for the 2PL IRT there was modest evidence of person- or item-level misfit using common diagnostic tools. The item difficulties were well-estimated, but the item discriminations were noticeably underestimated. As found by Kelderman and Molenaar, factor scores were estimated with less than expected reliability due to the underlying heterogeneity. The person specific models considered here are basically limiting cases of IRT models with multilevel, mixture, or differential item functioning structure. We conclude with some thoughts regarding real-world sources of heterogeneity that might go unacknowledged in common testing applications.

In psychology, the use of portable technology and wearable devices to ease participant burden in data collection is on the rise. This creates increased interest in collecting real-time or near real-time data from individuals within their natural environments. As a result, vast amounts of observational time series data are generated. Often, motivation for collecting this data hinges on understanding within-person processes that underlie psychological phenomena. Motivated by the body of Dr. Peter Molenaar's life work calling for analytical approaches that consider potential heterogeneity and non-ergodicity, the focus of this paper is on using idiographic analyses to generate population inferences for within-person processes. Meta-analysis techniques using one-stage and two-stage random effects meta-analysis as implemented in single-case experimental designs are presented. The case for preferring a two-stage approach for meta-analysis of single-subject observational time series data is made and demonstrated using an empirical example. This provides a novel implementation of the methodology as prior implementations focus on applications to short time series with experimental designs. Inspired by Dr. Molenaar's work, we describe how an approach, two-stage random effects meta-analysis (2SRE-MA), aligns with recent calls to consider idiographic approaches when making population-level inferences regarding within-person processes.

One type of genotype-environment interaction occurs when genetic effects on a phenotype are moderated by an environment; or when environmental effects on a phenotype are moderated by genes. Here we outline these types of genotype-environment interaction models, and propose a test of genotype-environment interaction based on the classical twin design, which includes observed genetic variables (polygenic scores: PGSs) that account for part of the genetic variance of the phenotype. We introduce environment-by-PGS interaction and the results of a simulation study to address statistical power and parameter recovery. Next, we apply the model to empirical data on anxiety and negative affect in children. The power to detect environment-by-PGS interaction depends on the heritability of the phenotype, and the strength of the PGS. The simulation results indicate that under realistic conditions of sample size, heritability and strength of the interaction, the environment-by-PGS model is a viable approach to detect genotype-environment interaction. In 7-year-old children, we defined two PGS based on the largest genetic association studies for 2 traits that are genetically correlated to childhood anxiety and negative affect, namely major depression (MDD) and intelligence (IQ). We find that common environmental influences on negative affect are amplified for children with a lower IQ-PGS.

Rapid developments over the last several decades have brought increased focus and attention to the role of time scales and heterogeneity in the modeling of human processes. To address these emerging questions, subgrouping methods developed in the discrete-time framework-such as the vector autoregression (VAR)-have undergone widespread development to identify shared nomothetic trends from idiographic modeling results. Given the dependence of VAR-based parameters on the measurement intervals of the data, we sought to clarify the strengths and limitations of these methods in recovering subgroup dynamics under different measurement intervals. Building on the work of Molenaar and collaborators for subgrouping individual time-series by means of the subgrouped chain graphical VAR (scgVAR) and the subgrouping option in the group iterative multiple model estimation (S-GIMME), we present results from a Monte Carlo study aimed at addressing the implications of identifying subgroups using these discrete-time methods when applied to continuous-time data. Results indicate that discrete-time subgrouping methods perform well at recovering true subgroups when the measurement intervals are large enough to capture the full range of a system's dynamics, either via lagged or contemporaneous effects. Further implications and limitations are discussed therein.

Self-regulating systems change along different timescales. Within a given week, a depressed person's affect might oscillate around a low equilibrium point. However, when the timeframe is expanded to capture the year during which they onboarded antidepressant medication, their equilibrium and oscillatory patterns might reorganize around a higher affective point. To simultaneously account for the meaningful change processes that happen at different time scales in complex self-regulatory systems, we propose a single model that combines a second-order linear differential equation for short timescale regulation and a first-order linear differential equation for long timescale adaptation of equilibrium. This model allows for individual-level moderation of short-timescale model parameters. The model is tested in a simulation study which shows that, surprisingly, the short and long timescales can fully overlap and the model still converges to the reasonable estimates. Finally, an application of this model to self-regulation of emotional well-being in recent widows is presented and discussed.

The cross-sectional correlation is frequently used to summarize psychological data, and can be considered the basis for many statistical techniques. However, the work of Peter Molenaar on ergodicity has raised concerns about the meaning and utility of this measure, especially when the interest is in discovering general laws that apply to (all) individuals. Through using Cattell's databox and adopting a multilevel perspective, this paper provides a closer look at the cross-sectional correlation, with the goal to better understand its meaning when ergodicity is absent. An analytical expression is presented that shows the cross-sectional correlation is a function of the between-person correlation (based on person-specific means), and the within-person correlation (based on individuals' temporal deviations from their person-specific means). Two curiosities related to this expression of the cross-sectional correlation are elaborated on, that is: a) the difference between the within-person correlation and the (average) person-specific correlation; and b) the unexpected scenarios that can arise because the cross-sectional correlation is a weighted sum rather than a weighted average of the between-person and within-person correlations. Seven specific examples are presented to illustrate various ways in which these two curiosities may combine; R code is provided, which allows researchers to investigate additional scenarios.

Despite previous warnings against the use of the difference-in-coefficients method for estimating the indirect effect when the outcome in the mediation model is binary, the difference-in-coefficients method remains readily used in a variety of fields. The continued use of this method is presumably because of the lack of awareness that this method conflates the indirect effect estimate and non-collapsibility. In this paper, we aim to demonstrate the problems associated with the difference-in-coefficients method for estimating indirect effects for mediation models with binary outcomes. We provide a formula that decomposes the difference-in-coefficients estimate into (1) an estimate of non-collapsibility, and (2) an indirect effect estimate. We use a simulation study and an empirical data example to illustrate the impact of non-collapsibility on the difference-in-coefficients estimate of the indirect effect. Further, we demonstrate the application of several alternative methods for estimating the indirect effect, including the product-of-coefficients method and regression-based causal mediation analysis. The results emphasize the importance of choosing a method for estimating the indirect effect that is not affected by non-collapsibility.

Among the most important merits of modern missing data techniques such as multiple imputation (MI) and full-information maximum likelihood estimation is the possibility to include additional information about the missingness process via auxiliary variables. During the past decade, the choice of auxiliary variables has been investigated under a variety of different conditions and more recent research points to the potentially biasing effect of certain auxiliary variables, particularly colliders (Thoemmes & Rose, 2014). In this article, we further extend biasing mechanisms of certain auxiliary variables considered in previous research and thereby focus on their effects on individual diagnosis based on norming, in which the whole distribution of a variable is of interest rather than average coefficients (e.g., means). For this, we first provide the theoretical underpinnings of the mechanisms under study and then provide two focused simulations that (i) directly expand on the collider scenario in Thoemmes and Rose (2014, appendix A) by considering outcomes that are relevant to norming and (ii) extend the scenarios under consideration by instrumental variable mechanisms. We illustrate the bias mechanisms for two different norming approaches and exemplify the procedures by means of an empirical example. We end by discussing limitations and implications of our research.

Researchers simulating covariance structure models sometimes add model error to their data to produce model misfit. Presently, the most popular methods for generating error-perturbed data are those by Tucker, Koopman, and Linn (TKL), Cudeck and Browne (CB), and Wu and Browne (WB). Although all of these methods include parameters that control the degree of model misfit, none can generate data that reproduce multiple fit indices. To address this issue, we describe a multiple-target TKL method that can generate error-perturbed data that will reproduce target RMSEA and CFI values either individually or together. To evaluate this method, we simulated error-perturbed correlation matrices for an array of factor analysis models using the multiple-target TKL method, the CB method, and the WB method. Our results indicated that the multiple-target TKL method produced solutions with RMSEA and CFI values that were closer to their target values than those of the alternative methods. Thus, the multiple-target TKL method should be a useful tool for researchers who wish to generate error-perturbed correlation matrices with a known degree of model error. All functions that are described in this work are available in the fungible $R$ library. Additional materials (e.g., $R$ code, supplemental results) are available at https://osf.io/vxr8d/.

To understand psychological data, it is crucial to examine the structure and dimensions of variables. In this study, we examined alternative estimation algorithms to the conventional GLASSO-based exploratory graph analysis (EGA) in network psychometric models to assess the dimensionality structure of the data. The study applied Bayesian conjugate or Jeffreys' priors to estimate the graphical structure and then used the Louvain community detection algorithm to partition and identify groups of nodes, which allowed the detection of the multi- and unidimensional factor structures. Monte Carlo simulations suggested that the two alternative Bayesian estimation algorithms had comparable or better performance when compared with the GLASSO-based EGA and conventional parallel analysis (PA). When estimating the multidimensional factor structure, the analytically based method (i.e., EGA.analytical) showed the best balance between accuracy and mean biased/absolute errors, with the highest accuracy tied with EGA but with the smallest errors. The sampling-based approach (EGA.sampling) yielded higher accuracy and smaller errors than PA; lower accuracy but also lower errors than EGA. Techniques from the two algorithms had more stable performance than EGA and PA across different data conditions. When estimating the unidimensional structure, the PA technique performed the best, followed closely by EGA, and then EGA.analytical and EGA.sampling. Furthermore, the study explored four full Bayesian techniques to assess dimensionality in network psychometrics. The results demonstrated superior performance when using Bayesian hypothesis testing or deriving posterior samples of graph structures under small sample sizes. The study recommends using the EGA.analytical technique as an alternative tool for assessing dimensionality and advocates for the usefulness of the EGA.sampling method as a valuable alternate technique. The findings also indicated encouraging results for extending the regularization-based network modeling EGA method to the Bayesian framework and discussed future directions in this line of work. The study illustrated the practical application of the techniques to two empirical examples in R.