Latent variable modeling

A dissertation combing both quantitative and qualitative analyses to describe the Old Assyrian social networks, the remains of a Middle Bronze Age (IIa) merchant colony with surviving documents from 1970 to 1720 B.C.E. including a hierarchical social reconstruction, chronological familial genealogies, and social network analysis of ca. 6,000 Old Assyrian texts from Kültepe, Turkey. The initial western discovery and subsequent excavations in the early 20th century and onwards yielded thousands of cuneiform tablets. By exploiting the names of the individual merchants and their relationships to each other, as recorded on the tablets, we demonstrate how to generate quantitative and qualitative social networks in order to populate them with hierarchical rank distributions reflecting the merchants’ age and seniority in the network, the results of a latent variable model described herein. A method for homonym disambiguation is delineated, and applied for the purpose of assessing the age of cohort groups for demographic analysis. We conclude that grounding the primary sources in a structural networked setting will allow for a closer examination of the textual narratives around events occurring in the trade-network scalable from multiple vantage points, from a single merchant’s household to the broadest extent of the trade networks, stretching 1,000 km between the city-state of Assur and this central trade hub in Anatolia, known as Kanesh.

Latent class analysis (LCA) and latent profile analysis (LPA) are powerful techniques that enable researchers to glean insights into “hidden” psychological experiences to create typologies and profiles to provide better-informed community-based policies and practice. These analytic methods have been used in a variety of domains, such as: psychosis symptomatology in the general population (Kibowski & Williams, 2012; Murphy, Shevlin, & Adamson, 2007; Shevlin, Murphy, Dorahy, & Adamson, 2007); substance abuse (Cleveland, Collins, Lanza, Greenberg, & Feinberg, 2010; James, McField, & Montgomery, 2013), peer victimization (Nylund, Bellmore, Nishina, & Graham, 2007), and anti-social/self-defeating behavior (Rosato & Baer, 2010). LCA and LPA are versatile methods of dealing with data of interest to community-based researchers in a deep and psychologically grounded way. This chapter will address the nuances of how and when to use LCA and LPA. Case studies of LCA and LPA will also be presented to illustrate the applicability of these techniques.

In this invited article, Cabrera-Nguyen provides guidelines for reporting scale development and validation results. Authors' attention to these guidelines will help ensure the research reported in JSSWR is rigorous and of high quality. This article provides guidance for those using exploratory factor analysis (EFA) and confirmatory factor analysis (CFA). In addition, the article provides helpful links to resources addressing structural equation modeling, multiple imputation for missing data, and a general resource for quantitative data analysis.

Trajectories of prosocial behavior and physical aggression between 6 and 12 years of age were identified for a sample (N=1,025) of males. The trajectories were then used to predict school dropout and physical violence at age 17. Using a group-based semi-parametric method, two trajectories of prosociality (low and moderate declining) and three trajectories of physical aggression (low, moderate, and high declining) were obtained. Only a small minority (3.4%) of the boys were characterized by both high aggression and moderate prosociality. Physical aggression predicted both school dropout and physical violence, but contrary to expectations, prosocial behavior did not have additive or protective effects.

Research on emotional labor focuses on how employees utilize 2 main regulation strategies—surface acting (i.e., faking one’s felt emotions) and deep acting (i.e., attempting to feel required emotions)—to adhere to emotional expectations of their jobs. To date, researchers largely have considered how each strategy functions to predict outcomes in isolation. However, this variable-centered perspective ignores the possibility that there are subpopulations of employees who may differ in their combined use of surface and deep acting. To address this issue, we conducted 2 studies that examined surface acting and deep acting from a person-centered perspective. Using latent profile analysis, we identified 5 emotional labor profiles—non-actors, low actors, surface actors, deep actors, and regulators—and found that these actor profiles were distinguished by several emotional labor antecedents (positive affectivity, negative affectivity, display rules, customer orientation, and emotion demands–abilities fit) and differentially predicted employee outcomes (emotional exhaustion, job satisfaction, and felt inauthenticity). Our results reveal new insights into the nature of emotion regulation in emotional labor contexts and how different employees may characteristically use distinct combinations of emotion regulation strategies to manage their emotional expressions at work.

Testing for measurement invariance can be done within the context of multigroup latent class analysis. Latent class analysis can model any type of discrete level data, which makes it an obvious choice when nominal indicators are used or when a researcher's aim is at classifying respondents in latent classes. The multigroup latent class (LC) model can be specified in three different ways, i.e. by adopting a probabilistic, a log-linear or a logistic parameterization. We define and compare these different forms of parameterization. The starting point is the standard LC model in which indicators and latent variables are defined at the nominal level. Additionally, we focus on LC models with ordinal indicators as well as LC factor models with ordinal indicators. Testing for measurement invariance involves estimating LC models with different degrees of homogeneity. We explain the procedure for investigating measurement invariance at both the scale as well as the item level. We illustrate the approach with two examples. The first example is a multigroup LC analysis with nominal indicators; the second a multigroup LC factor analysis with ordinal indicators.

Les fouilles récentes ont révélé la topographie des sites agglomérés de la région toulousaine aux IIe et Ier s. av. n. ère. Pour la première fois a été mise en évidence la trame urbaine de l'oppidum de Vieille-Toulouse, trame apparue entre 150 et 125 et maintenue, avec des ajustements, jusqu'à l'abandon du site vers - 10. Dans la plaine à Saint-Roch, l'occupation gauloise couvre une surface de l'ordre de 80 ha, sans structuration de grande ampleur. Il s'agit d'une agglomération secondaire sans doute dédiée à des activités spécifiques liées à la Garonne.

Simultaneous latent-class analysis across groups (SLCAG) is an extension of the standard latent class (LC) model for the examination of measurement equivalence/invariance. It can be used to compare the latent structure derived from a set of discrete item responses between multiple groups, e.g., between males and females, between Japanese, Americans and Dutch, young and old, and between ill and healthy. Whereas the more commonly used multigroup confirmatory factor analysis (MCFA) assumes that the underlying latent variables are continuous, SLCAG treats the latent variables either as nominal – e.g. to identify a typological classification from a given set of categorical indicators – or as ordinal – e.g. to investigate the scalability of a set of categorical indicators. These two specifications are sometimes referred as LC cluster and LC factor models, respectively.

Being part of a delinquent group has been shown to facilitate the expression of an individual's own delinquent propensities. However, this facilitation effect has not been investigated from a developmental perspective within a population heterogeneity model. Using a semiparametric mixture model with data from the Montreal Longitudinal Experimental Study, this article addresses important issues in the developmental trends of membership to delinquent groups. We explore how the rate of violent behaviors follows delinquent peer group trajectories and investigate a differential facilitation effect of delinquent peers on violence across multiple developmental pathways. Results suggest that 25% of males followed a childhood or an adolescence delinquent group affiliation trajectory. These two groups account far most of the violent acts assessed during adolescence. In addition, the rate of violent behaviors follows these developmental trajectories. Controlling for these delinquent group trajectories, we also found that being involved in a delinquent group at any specific time during adolescence is associated with an increased rate of violent behaviors, and that leaving these groups results in a decrease in violent behaviors. This facilitation effect appears homogeneous over time and across developmental trajectories. Results are discussed from a social interactional perspective.

A Monte Carlo approach was used to examine bias in the estimation of indirect effects and their associated standard errors. In the simulation design, (a) sample size, (b) the level of nonnormality characterizing the data, (c) the population values of the model parameters, and (d) the type of estimator were systematically varied. Estimates of model parameters were generally unaffected by either nonnormality or small sample size. Under severely nonnormal conditions, normal theory maximum likelihood estimates of the standard error of the mediated effect exhibited less bias (approximately 10% to 20% too small) compared to the standard errors of the structural regression coefficients (20% to 45% too small). Asymptotically distribution free standard errors of both the mediated effect and the structural parameters were substantially affected by sample size, but not nonnormality. Robust standard errors consistently yielded the most accurate estimates of sampling variability.

This research examines empathic dispositions of 178 pre-service teachers. We analyzed open ended responses to animated narrative vignette simulations (ANVs), which served as stimulated
experimental situations depicting students in victim and perpetrator scenarios. Empathy was examined by addressing the following questions: (1) Do participants’ responses differ over vignettes? (2) What is the dimensionality of the empathy construct? (3) Is word count an indicator of empathy?
(4) Is there a dispositional effect? (5) To what extent do pre-service teachers express empathy? After the text responses of pre-service teachers were coded, log-linear and log-multiplicative association
models, which have graphical representations, were used to analyze the data and to develop a context dependent measure of empathy. The results suggest that a single latent variable underlies the responses, and from our measurement model, very few teachers expressed empathy toward the victim in the ANVs.

The purpose of this study was to investigate (a) the latent profiles that arise from middle and high school students' (N = 1225) reported exposure to information from the four hypothesized sources of self-efficacy; (b) the relationships between these latent profiles and science self-efficacy and science achievement; and (c) the differences in latent profiles as a function of implicit theory of science ability, gender, and grade level. Results revealed that a four-class solution fit the data the best. Results support past findings indicating that mastery experiences are a powerful source of self-efficacy. Furthermore, there seemed to be an additive benefit of drawing from multiple sources simultaneously. Gender did not predict membership in these four profiles, but implicit theory of ability and grade level did. The results show that students in the most adaptive profiles drew from multiple sources of efficacy-relevant information and espoused a strong belief in the plasticity of their science abilities, whereas those who were in the least adaptive profiles exhibited a high degree of negative affect and held a fixed view of science ability.

A common situation in the evaluation of intervention programs is the researcher’s possibility to rely on two waves of data only (i.e., pretest and posttest), which profoundly impacts on his/her choice about the possible statistical analyses to be conducted. Indeed, the evaluation of intervention programs based on a pretest-posttest design has been usually carried out by using classic statistical tests, such as family-wise ANOVA analyses, which are strongly limited by exclusively analyzing the intervention effects at the group level. In this article, we showed how second order multiple group latent change modeling (SO-MG-LCM) could represent a useful methodological tool to have a more realistic and informative assessment of intervention programs with two waves of data. We offered a practical step-by-step guide to properly implement this methodology, and we outlined the advantages of the LCM approach over classic ANOVA analyses. Furthermore, we also provided a real-data example by re-analyzing the implementation of the Young Prosocial Animation, a universal intervention program aimed at promoting prosociality among youth. In conclusion, albeit there are previous studies that pointed to the usefulness of MG-LCM to evaluate intervention programs (Curran & Muthén, 1999; Muthén & Curran, 1997), no previous study showed that it is possible to use this approach even in pretest-posttest (i.e., with only two time points) designs. Given the advantages of latent variable analyses in examining differences in interindividual and intraindividual changes (McArdle, 2009), the methodological and substantive implications of our proposed approach are discussed.

The network approach to psychopathology is becoming increasingly popular. The motivation for this approach is to provide a replacement for the problematic common cause perspective and the associated latent variable model, where symptoms are taken to be mere effects of a common cause (the disorder itself). The idea is that the latent variable model is plausible for medical diseases, but unrealistic for mental disorders, which should rather be conceptualized as networks of directly interacting symptoms.  We argue that this rationale for the network approach is misguided. Latent variable (or common cause) models are not inherently problematic, and there is not even a clear boundary where network models end and latent variable (or common cause) models begin. We also argue that focusing on this contrast has led to an unrealistic view of testing and finding support for the network approach, as well as an oversimplified picture of the relationship between medical diseases and mental disorders.  As an alternative, we point out more essential contrasts, such as the contrast between dynamic and static modeling approaches, that can provide a better framework for conceptualizing mental disorders. Finally, we discuss several topics and open problems that need to be addressed in order to make the network approach more concrete and to move the field of psychological network research forward.

The purpose of this study was to explore (a) the individual belief profiles that naturally arise among middle and high school science students (n = 1225); (b) the relationships between these profiles to science achievement and other prominent motivation variables; and (c) the demographic and developmental differences among the belief profiles. Results revealed that a four-class solution fit the data the best. These profiles were differentially related to achievement goal orientations, science self-efficacy, and science achievement. Differences in profiles also arose as a function of minority status, grade level, and gender. Findings support and refine Schommer-Aikins's (2004) Embedded Systemic Model of epistemic beliefs. Results are discussed in relation to theory and implications for science instruction.

Background: Different developmental courses have been postulated for proactive and reactive aggression. Objective: Investigated the developmental course of proactive and reactive aggression in a large sample of adolescent boys from low socioeconomic areas. Method: A dual group-based joint trajectory method was used to identify distinct trajectories as well as similarities and differences in intra-individual changes. Results: The trajectories for proactive and reactive aggression were similar: the majority of individuals followed infrequent and desisting trajectories. Contrary to expectations, very few adolescents followed trajectories of increasing proactive aggression. Reactive aggression was more common than proactive aggression. The overlap in trajectory group membership of individuals following trajectories of high peaking proactive and reactive aggression was nearly 100%. Across a period of 5 years, the boys on the high peaking trajectories were twice as likely to have affiliated with gangs. Conclusions: The developmental courses of proactive and reactive aggression are similar during adolescence. Males who tend to frequently use one form of aggression throughout adolescence also tend to frequently use the other and are at an increased risk for contemporaneous delinquent lifestyles.

This paper presents a software package designed to estimate Poole and Rosenthal W-NOMINATE scores in R. The package uses a logistic regression model to analyze political choice data, usually (though not exclusively) from a legislative setting. In contrast to other scaling methods, W-NOMINATE explicitly assumes probabilistic voting based on a spatial utility function, where the parameters of the utility function and the spatial coordinates of the legislators and the votes can all be estimated on the basis of observed voting behavior. Building on software written by Poole in Fortran, the new wnominate package in R facilitates easier data input and manipulation, generates bootstrapped standard errors, and includes a new suite of graphics functions to display the results. We demonstrate the functionality of this package by conducting a natural experiment using roll callsan experiment which is greatly simplified by the data manipulation capabilities of the wnominate package in R.

There are many arguments for and against the use of autonomous-agents in intelligent environments. Some researchers maintain that it is of utmost importance to give complete control to users, and hence greatly restrict autonomy of agents; whereas, others believe that is it preferable to increase user convenience by allowing agents to operate autonomously on the user’s behalf. While both of these approaches have their distinct merits, they are not suitable for all users. As people’s opinions and concerns regarding agent autonomy are highly individual, depending on a wide range of factors and often changing over time, a much more dynamic approach to agent autonomy is needed. This work explores how it is possible to equip intelligent environments with an adjustable autonomy mechanism, which allows an individual user to increase or decrease agent autonomy in order to find their own comfortable sweet-spot between maintaining/relinquishing control and gaining/losing convenience. This paper presents the Adjustable Autonomy Intelligent Environment (AAIE) model, discusses how adjustable autonomy can be achieved in intelligent environments, and discusses the major findings from a recent online survey and user study, which highlight the major factors and concerns of users that determine their personal preferences towards different levels of autonomy.

This article presents a combined motivational and volitional intervention based on the theory of planned behavior aimed at promoting expansion-oriented job crafting behaviors. Participants were employees working in different companies, assigned to either an intervention (n = 53) or a control group (n = 55). Results of a field study (including premeasure, postmeasure, and weekly diaries) indicated that the intervention enhanced participants’ perceptions of behavioral control referred to job crafting and awareness regarding others’ engagement in job crafting. Latent change growth modeling showed that participation in the intervention led to participants shaping their job crafting intentions during the weeks, which translated into more frequent job crafting behaviors at the end of the study period. Besides, the intervention served to trigger weekly work-related flow experiences in terms of high absorption while working. Findings suggest that job crafting interventions can benefit from the inclusion of self-regulatory strategies complementing goal setting.

This study examined the structure of posttraumatic stress disorder (PTSD) as measured by the Impact of Event Scale-Revised (IES-R; Weiss & Marmar, 1997), tested factorial invariance for samples of 235 Israeli emergency room patients and 306 U.S. undergraduate students, and then evaluated factorial invariance over multiple occasions within the emergency room sample. A four-factor structure representing intrusion, avoidance-numbing, hyperarousal, and sleep emerged as the preferred model. Configural invariance over groups was supported for this model. Likewise, configural invariance over occasions was demonstrated, but metric invariance was not fully supported, with variation in the loadings on the intrusion factor over time seemingly the source of misfit. Interpretations and conclusions center on sleep as a separate factor underlying the structure of the IES-R, the distinction between avoidance and numbing as a function of how the IES-R (vs. the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition) operationalizes the numbing feature of PTSD and possible shifts in the meaning of intrusion over time.

We provide a comprehensive review of simple and advanced statistical analyses using an intuitive visual approach explicitly modeling Latent Variables (LV). This method can better illuminate what is assumed in each analytical method and what is actually estimated, by translating the causal relationships embedded in the graphical models in equation form. We recommend the graphical display rooted in the century old
path analysis, that details all parameters of each statistical model, and suggest labeling that clarifies what is given vs. what is estimated.
We link in the process classical and modern analyses under the
encompassing broader umbrella of Generalized Latent Variable
Modeling, and demonstrate that LVs are omnipresent in all statistical
approaches, yet until directly ‘seeing’ them in visual graphical displays, they are unnecessarily overlooked. The advantages of directly modeling LVs are shown with examples of analyses from the Active8 intervention designed to increase physical activity.

A longitudinal study with a nested preventive intervention was used to test five hypotheses generated from developmental theories of antisocial behavior. The longitudinal study followed 909 boys from their kindergarten year up to 17 years of age. The randomized multimodal preventive intervention targeted a subsample of boys who were rated disruptive by their kindergarten teacher. Semiparametric analyses of developmental trajectories for self-reported physical aggression, vandalism, and theft identified more types of trajectories than expected from recent theoretical models. Also, these trajectories did not confirm theoretical models, which suggest a general increase of antisocial behavior during adolescence. The majority of boys were on either a low-level antisocial behavior trajectory or a declining trajectory. Less than 6% appeared to follow a trajectory of chronic antisocial behavior. Comparisons between disruptive and nondisruptive kindergarten boys confirmed the hypothesis that disruptive preschool children are at higher risk of following trajectories of frequent antisocial behavior. Comparisons between treated and untreated disruptive boys confirmed that an intensive preventive intervention between 7 and 9 years of age, which included parent training and social skills training, could change the long-term developmental trajectories of physical aggression, vandalism, and theft for disruptive kindergarten boys in low socioeconomic areas. The results suggest that trajectories of violent behavior can be deflected by interventions that do not specifically target the physiological deficits that are often hypothesized to be a causal factor. The value of longitudinal-experimental studies from early childhood onward is discussed.

Brain imaging is increasingly recognised as an intermediate phenotype to understand the complex path between genetics and behavioural or clinical phenotypes. In this context, a first goal is to propose methods to identify the part of genetic variability that explains some neuroimaging variability. Classical univariate approaches often ignore the potential joint effects that may exist between genes or the potential covariations between brain regions. In this paper, we propose instead to investigate an exploratory multivariate method in order to identify a set of Single Nucleotide Polymorphisms (SNPs) covarying with a set of neuroimaging phenotypes derived from functional Magnetic Resonance Imaging (fMRI). Recently, Partial Least Squares (PLS) regression or Canonical Correlation Analysis (CCA) have been proposed to analyse DNA and transcriptomics. Here, we propose to transpose this idea to the DNA vs. imaging context. However, in very high-dimensional settings like in imaging genetics studies, such multivariate methods may encounter overfitting issues. Thus we investigate the use of different strategies of regularisation and dimension reduction techniques combined with PLS or CCA to face the very high dimensionality of imaging genetics studies. We propose a comparison study of the different strategies on a simulated dataset first and then on a real dataset composed of 94 subjects, around 600,000 SNPs and 34 functional MRI lateralisation indexes computed from reading and speech comprehension contrast maps. We estimate the generalisability of the multivariate association with a cross-validation scheme and demonstrate the significance of this link, using a permutation procedure. Univariate selection appears to be necessary to reduce the dimensionality. However, the significant association uncovered by this two-step approach combining univariate filtering and L1regularised PLS suggests that discovering meaningful genetic associations calls for a multivariate approach.

The cross-classified multiple membership latent variable regression (CCMM-LVR) model is a recent extension to the three-level latent variable regression (HM3-LVR) model which can be utilized for longitudinal data that contains individuals who changed clusters over time (for instance, student mobility across schools). The HM3-LVR model can include the initial status on growth effect as varying across those clusters and allows testing of more flexible hypotheses about the influence of initial status on growth and of factors that might impact that relationship, but only in the presence of pure clustering of participants within higher-level units. This Monte Carlo study was conducted to evaluate model estimation under a variety of conditions and to measure the impact of ignoring cross-classified data when estimating the incorrectly specified HM3-LVR model in a scenario in which true values for parameters are known. Furthermore, results from a real-data analysis were used to inform the design of the simulation. Overall, it would be recommended for researchers to utilize the CCMM-LVR model over the HM3-LVR model when individuals are cross-classified, and to use a bare minimum of more than 100 clustering units in order to avoid overestimation of the level-3 variance component estimates.

Correlation filters are special classifiers designed for shift-invariant object recognition, which are robust to pattern distortions. The recent literature shows that combining a set of sub-filters trained based on a single or a small group of images obtains the best performance. The idea is equivalent to estimating variable distribution based on the data sampling (bagging), which can be interpreted as finding solutions (variable distribution approximation) directly from sampled data space. However, this methodology fails to account for the variations existed in the data. In this paper, we introduce an intermediate step – solution sampling – after the data sampling step to form a subspace, in which an optimal solution can be estimated. More specifically, we propose a new method, named latent constrained correlation filters (LCCF), by mapping the correlation filters to a given latent subspace, and develop a new learning framework in the latent subspace that embeds distribution-related constraints into the original problem. To solve the optimization problem, we introduce a subspace based alternating direction method of multipliers (SADMM), which is proven to converge at the saddle point. Our approach is successfully applied to three different tasks, including eye localization, car detection and object tracking. Extensive experiments demonstrate that LCCF outperforms the state-of-the-art methods. 1

consistently high likelihood and use frequency; and (iii) high likelihood and moderate use frequency. Covariate analyses revealed improvement in PTSD severity was associated with membership in a specific substance use trajectory, although receiving trauma-informed treatment was not. Additionally, SUD severity, age and after-care efforts were shown to be related to trajectory membership. Discussion and Conclusions. Findings highlight the necessity of accounting for heterogeneity in post-treatment substance use, relevance of trauma-informed care in SUD recovery and benefits of incorporating methodologies like LGMM when evaluating SUD treatment outcomes. [López-Castro T, Hu M-C, Papini S, Ruglass LM, Hien DA. Pathways to change: Use trajectories following trauma-informed treatment of women with co-occurring posttraumatic stress disorder and substance use disorders. Drug Alcohol Rev 2015]

The variance in satisfaction with life can be broken down into trait- and state-like components.Weran tests to determine if a new scale for the measurement of satisfaction with life, the Steen Happiness Index (SHI), was more sensitive to situational changes than the Satisfaction with Life Scale (SWLS), which is currently used as a gold standard, with wellestablished state–trait characteristics. The study consisted of 292 young adults aged 18–35, who were recruited and interviewed three times with 5-month intervals at their places of residence. They completed a set of questionnaires involving the SHI and the SWLS. A latent autoregressive state–trait model showed an adequate fit, v2 (56) = 81.35, p = .02, v2/df = 1.45, RMSEA = .06, CFI = .98, SRMR = .06. The results indicated that the two scales are strongly correlated on the trait level (r = .98) and on the occasion-specific level (between r = .53 and r = .65) variables. For both scales, stable and occasion-specific influences determine inter-individual differences. For both scales, it is not possible to predict half of the variance by either traits or states measured earlier in time.

O objetivo deste artigo é o desenvolvimento de um modelo de mensuração dos sentimentos de legitimidade policial da população paulistana. Identificamos que, na literatura nacional a respeito das atitudes públicas sobre as autoridades legais, pesquisas que se valem de desenhos de survey comumente adotam uma perspectiva operacionalista, negligenciando a lacuna existente entre os conceitos e as questões. Contribuímos substantivamente para esse campo desenvolvendo um modelo de mensuração de legitimidade policial, definida conceitualmente como um dever normativamente orientado de obedecer e um alinhamento moral com a polícia. Utilizando dados de um survey representativo de oito regiões da cidade de São Paulo em 2017, realizamos uma análise fatorial confirmatória de segunda ordem. Posteriormente,validamos essa medida por meio de um modelo de equações estruturais centrado no construto legitimidade. Implicações do conceito e da mensuração de legitimidade para a segurança pública são discutidas.

mean, min and max intercorrelations between the subscales of the SCL-90 and SCL-90-R from eight different studies. Studies who are listed with more than one set of data reported data from more than one sample representing distinct populations (e.g. Schmitz et al. (2000) examined psychosomatic outpatients and unscreened primary care patients). The mean correlations reported across studies represent moderate to high correlations between supposedly independent symptom dimensions. These high intercorrelations represent a significant threat to the construct validity of the SCL-90 and SCL-90-R as multidimensional instruments.

CONTEXT:
Deviant peer group involvement is strongly related to onset, aggravation, and persistence of conduct problems during adolescence.
OBJECTIVE:
To identify early childhood behavioral profiles that predict early-onset deviant peer group involvement.

Study of multivariate data in situations where a variable of interest
is unobservable (latent) and only measured indirectly is widely applied. Item
response models are powerful tools for measurement and have been extended
to incorporate latent structure. The (loglinear) Rasch model is a simple item
response model where tests of fit and item parameters estimation can take place
without assumptions about the distribution of the latent variable. Inclusion of a
latent variable as predictor in standard regression models like logistic or Poisson
regression models is discussed, and a study of the relation between psychosocial
work environment and absence from work is used to illustrate and motivate the
results.

Understanding the mental state of other people is an important skill for intelligent agents and robots to operate within social environments. However, the mental processes involved in ‘mind-reading’ are complex. One explanation of such processes is Simulation Theory — it is supported by a large body of neuropsychological research. Yet, determining the best computational model or theory to use in simulation-style emotion detection, is far from being understood. In this work, we use Simulation Theory and neuroscience findings on Mirror-Neuron Systems as the basis for a novel computational model, as a way to handle affective facial expressions. The model is based on a probabilistic mapping of observations from multiple identities onto a single fixed identity (‘internal transcoding of external stimuli’), and then onto a latent space (‘phenomenological response’). Together with the proposed architecture we present some promising preliminary results.

In this paper, a novel technique for blind signal separation based on a combination of second order and higher order approaches is introduced. The problem of blind signal separation was solved in a wavelet domain. The main idea behind this approach is that the mixing signal can be decomposed into a sum of uncorrelated and/or independent sub-bands using the wavelet transform. In the beginning, the observed signal is prewhitened in the time domain then, the initial separation matrix will be estimated from second order statistics decorrelation method in the wavelet domain. The estimating matrix will be used as an initial separating matrix in the higher order statistics method in order to estimate the final separation matrix. The algorithm was tested using natural images. Extensive experiments have confirmed that the use of the proposed procedure provides promising results in separating the image from noisy mixtures of images.

In this paper, we propose a Gaussian process (GP) model for analysis of nonlinear time series. Formulation of our model is based on the consideration that the observed data are functions of latent variables, with the associated mapping between observations and latent representations modeled through GP priors. In addition, to capture the temporal dynamics in the modeled data, we assume that subsequent latent representations depend on each other on the basis of a hidden Markov prior imposed over them. Derivation of our model is performed by marginalizing out the model parameters in closed form using GP priors for observation mappings, and appropriate stick- breaking priors for the latent variable (Markovian) dynamics. This way, we eventually obtain a nonparametric Bayesian model for dynamical systems that accounts for uncertainty in the modeled data. We provide efficient inference algorithms for our model on the basis of a truncated variational Bayesian approximation. We demonstrate the efficacy of our approach considering a number of applications dealing with real-world data, and compare it with the related state-of-the-art approaches.

According to cognitive load theory, instructions can impose three types of cognitive load on the learner: intrinsic load, extraneous load, and germane load. Proper measurement of the different types of cognitive load can help us understand why the effectiveness and efficiency of learning environments may differ as a function of instructional formats and learner characteristics. In this article, we present a ten-item instrument for the measurement of the three types of cognitive load. Principal component analysis on data from a lecture in statistics for PhD students (n = 56) in psychology and health sciences revealed a threecomponent solution, consistent with the types of load that the different items were intended to measure. This solution was confirmed by a confirmatory factor analysis of data from three lectures in statistics for different cohorts of bachelor students in the social and health sciences (ns = 171, 136, and 148), and received further support from a randomized experiment with university freshmen in the health sciences (n = 58).

Originating from a system theory and an input/output point of view, I introduce a new class of generalized distributions. A parametric nonlinear transformation converts a random variable X into a so-called Lambert W random variable Y, which allows a very flexible approach to model skewed data. Its shape depends on the shape of X and a skewness parameter γ. In particular, for symmetric X and nonzero γ the output Y is skewed. Its distribution and density function are particular variants of their input counterparts. Maximum likelihood and method of moments estimators are presented, and simulations show that in the symmetric case additional estimation of γ does not affect the quality of other parameter estimates. Applications in finance and biomedicine show the relevance of this class of distributions, which is particularly useful for slightly skewed data. A practical by-result of the Lambert W framework: data can be “unskewed.”

The R package LambertW developed by the author is publicly available (CRAN).

This article shows how interfactor correlation is affected by error correlations. Theoretical and practical justifications for error correlations are given, and a new equivalence class of models is presented to explain the relationship between interfactor correlation and error correlations. The class allows simple, parsimonious modeling of error correlations via prespecifying reliabilities. Within the class, the correlation between latent factors can be as high as 1.0, and as low as
the correlation between certain component scores. The models are indistinguishable in terms of parameter parsimony, identifiability, and fit statistics, implying that interfactor correlation is not identifiable within the class. The existence of the class is problematic for psychometric measurement, because estimates of interfactor correlation form the foundation of much of the literature.

The major challenge of learning from multi-label data has arisen from the overwhelming size of label space which makes this problem NP-hard. This problem can be alleviated by gradually involving easy to hard tags into the learning process. Besides, the utilization of a diversity maintenance approach avoids overfitting on a subset of easy labels. In this paper, we propose a self-paced multi-label learning with diversity (SPMLD) which aims to cover diverse labels with respect to its learning pace. In addition, the proposed framework is applied to an efficient correlation-based multi-label method. The non-convex objective function is optimized by an extension of the block coordinate descent algorithm. Empirical evaluations on real-world datasets with different dimensions of features and labels imply the effectiveness of the proposed predictive model.

Performing actions in a timely manner is an indispensable aspect in everyday human activities. Accordingly, it has to be present in robotic systems if they are going to seamlessly interact with humans. The current work addresses the problem of learning both the spatial and temporal characteristics of human motions from observation. We formulate learning as a mapping between two worlds (the observed and the action ones). This mapping is realized via an abstract intermediate representation termed "Latent Space." Learned actions can be subsequently invoked in the context of more complex human-robot interaction (HRI) scenarios. Unlike previous learning from demonstration (LfD) methods that cope only with the spatial features of an action, the formulated scheme effectively encompasses spatial and temporal aspects. Learned actions are reproduced under the high-level control of a time-informed task planner. During the implementation of the studied scenarios, temporal and physical constraints may impose speed adaptations in the reproduced actions. The employed latent space representation readily supports such variations, giving rise to novel actions in the temporal domain. Experimental results demonstrate the effectiveness of the proposed scheme in the implementation of HRI scenarios. Finally, a set of well-defined evaluation metrics are introduced to assess the validity of the proposed approach considering the temporal and spatial consistency of the reproduced behaviors.

The study of temporal dynamics is essential to the advance of social science. In the study of inequality, preliminary to explaining patterns in gaps between groups is the prior task of detecting those patterns. Developing a multiple latent trajectory strategy, this paper proposes an inductive approach to the detection of inequality trends. Investigating managerial representation for Black men, Black women, White women and White men in a very large panel sample of private sector workplaces, we show that trajectories of managerial representation for each status group are much more complex than the common deductive approach leads us to believe. In fact, the volatility and the multiplicity of trajectories is one of the markers of enduring disadvantage for African Americans. Further, we demonstrate that current organizational inequality explanations can be employed and improved within this inductive methodology. The findings underscore the importance of interaction and contingency in the production of inequalities, in agreement with relational approaches to organizational inequality.