Social network data represent interactions and relationships among groups of individuals. One aspect of social interaction is social influence, the idea that beliefs or behaviors change as a result of one’s social network. The purpose of this article is to introduce a new model for social influence, the latent space model for influence, which employs latent space positions so that individuals are affected most by those who are “closest” to them in the latent space. We describe this model along with some of the contexts in which it can be used and explore the operating characteristics using a series of simulation studies. We conclude with an example of teacher advice-seeking networks to show that changes in beliefs about teaching mathematics may be attributed to network influence.

A maximum likelihood estimation routine for two-level structural equation models with random slopes for latent covariates is presented. Because the likelihood function does not typically have a closed-form solution, numerical integration over the random effects is required. The routine relies upon a method proposed by du Toit and Cudeck (Psychometrika 74(1):65–82, 2009) for reformulating the likelihood function so that an often large subset of the random effects can be integrated analytically, reducing the computational burden of high-dimensional numerical integration. The method is demonstrated and assessed using a small-scale simulation study and an empirical example.

A shadow-test approach to the calibration of field-test items embedded in adaptive testing is presented. The objective function used in the shadow-test model selects both the operational and field-test items adaptively using a Bayesian version of the criterion of $D_{\mathrm{s}}$\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$D_{\mathrm{s}}$$\end{document}-optimality. The constraint set for the model can be used to hide the field-test items completely in the content of the test as well as to deal with such practical issues as random control of their exposure rates. The approach runs on efficient implementations of the Gibbs sampler for the real-time updating of the ability and field-test parameters. Optimal settings for the proposed algorithms were found and used to demonstrate item calibration with smaller than traditional sample sizes in runtimes fully comparable with conventional adaptive testing.

In diagnostic classification models (DCMs), the Q matrix encodes in which attributes are required for each item. The Q matrix is usually predetermined by the researcher but may in practice be misspecified which yields incorrect statistical inference. Instead of using a predetermined Q matrix, it is possible to estimate it simultaneously with the item and structural parameters of the DCM. Unfortunately, current methods are computationally intensive when there are many attributes and items. In addition, the identification constraints necessary for DCMs are not always enforced in the estimation algorithms which can lead to non-identified models being considered. We address these problems by simultaneously estimating the item, structural and Q matrix parameters of the Deterministic Input Noisy “And” gate model using a constrained Metropolis–Hastings Robbins–Monro algorithm. Simulations show that the new method is computationally efficient and can outperform previously proposed Bayesian Markov chain Monte-Carlo algorithms in terms of Q matrix recovery, and item and structural parameter estimation. We also illustrate our approach using Tatsuoka’s fraction–subtraction data and Certificate of Proficiency in English data.

We revisit a singular value decomposition (SVD) algorithm given in Chen et al. (Psychometrika 84:124–146, 2019b) for exploratory item factor analysis (IFA). This algorithm estimates a multidimensional IFA model by SVD and was used to obtain a starting point for joint maximum likelihood estimation in Chen et al. (2019b). Thanks to the analytic and computational properties of SVD, this algorithm guarantees a unique solution and has computational advantage over other exploratory IFA methods. Its computational advantage becomes significant when the numbers of respondents, items, and factors are all large. This algorithm can be viewed as a generalization of principal component analysis to binary data. In this note, we provide the statistical underpinning of the algorithm. In particular, we show its statistical consistency under the same double asymptotic setting as in Chen et al. (2019b). We also demonstrate how this algorithm provides a scree plot for investigating the number of factors and provide its asymptotic theory. Further extensions of the algorithm are discussed. Finally, simulation studies suggest that the algorithm has good finite sample performance.

Computer-based interactive items have become prevalent in recent educational assessments. In such items, detailed human–computer interactive process, known as response process, is recorded in a log file. The recorded response processes provide great opportunities to understand individuals’ problem solving processes. However, difficulties exist in analyzing these data as they are high-dimensional sequences in a nonstandard format. This paper aims at extracting useful information from response processes. In particular, we consider an exploratory analysis that extracts latent variables from process data through a multidimensional scaling framework. A dissimilarity measure is described to quantify the discrepancy between two response processes. The proposed method is applied to both simulated data and real process data from 14 PSTRE items in PIAAC 2012. A prediction procedure is used to examine the information contained in the extracted latent variables. We find that the extracted latent variables preserve a substantial amount of information in the process and have reasonable interpretability. We also empirically prove that process data contains more information than classic binary item responses in terms of out-of-sample prediction of many variables.

Theory of mind (ToM) is an essential social-cognitive ability to understand one’s own and other people’s mental states. Neural data as well as behavior data have been utilized in ToM research, but the two types of data have rarely been analyzed together, creating a large gap in the literature. In this paper, we propose and apply a novel joint modeling approach to analyze brain activations with two types of behavioral data, response times and response accuracy, obtained from a multi-item ToM assessment, with the intention to shed new light on the nature of the underlying process of ToM reasoning. Our trivariate data analysis suggested that different levels or kinds of processes might be involved during the ToM assessment, which seem to differ in terms of cognitive efficiency and sensitivity to ToM items and the correctness of item responses. Additional details on the trivariate data analysis results are provided with discussions on their implications for ToM research.

There has been regained interest in joint maximum likelihood (JML) estimation of item factor analysis (IFA) recently, primarily due to its efficiency in handling high-dimensional data and numerous latent factors. It has been established under mild assumptions that the JML estimator is consistent as both the numbers of respondents and items tend to infinity. The current work presents an efficient Riemannian optimization algorithm for JML estimation of exploratory IFA with dichotomous response data, which takes advantage of the differential geometry of the fixed-rank matrix manifold. The proposed algorithm takes substantially less time to converge than a benchmark method that alternates between gradient ascent steps for person and item parameters. The performance of the proposed algorithm in the recovery of latent dimensionality, response probabilities, item parameters, and factor scores is evaluated via simulations.

Complex problem solving (CPS) is an up-and-coming twenty-first century skill that requires test-takers to solve dynamically changing problems, often assessed using computer-based tests. The log data that users produce when interacting with a computer-based test provide valuable information about each individual behavioral action they undertake, but such data are rather difficult to handle from a statistical point of view. This paper addresses this issue by building upon recent research focused on decoding log data and aims to identify homogeneous student profiles with regard to their ability to solve CPS tasks. Therefore, we estimated a discrete two-tier item response theory model, which allowed us to profile units (i.e., students) while taking into account the multidimensionality of the data and the explanatory effect of individual characteristics. The results indicate that: (1) CPS can be thought of as a three-dimensional latent variable; (2) there are ten latent classes of students with homogenous profiles regarding the CPS dimensions; (3) students in the higher latent classes generally demonstrate higher cognitive and non-cognitive performances; (4) some of the latent classes seem to profit from learning-by-doing within tasks, whereas others seem to exhibit the reverse behavior; (5) cognitive and non-cognitive skills, as well as gender and to some extent age, contribute to distinguishing among the latent classes.

We generalize the power divergence (PD) family of statistics to the two-parameter logistic IRT model for the purpose of constructing hypothesis tests and confidence intervals of the person parameter. The well-known score test statistic is a special case of the proposed PD family. We also prove the proposed PD statistics are asymptotically equivalent and converge in distribution to ${\chi }_1^2$\documentclass[12pt]{minimal}\usepackage{amsmath}\usepackage{wasysym}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{amsbsy}\usepackage{mathrsfs}\usepackage{upgreek}\setlength{\oddsidemargin}{-69pt}\begin{document}$$\chi _{1}^2$$\end{document}. In addition, a moment matching method is introduced to compare statistics and choose the optimal one within the PD family. Simulation results suggest that the coverage rate of the associated confidence interval is well controlled even under small sample sizes for some PD statistics. Compared to some other approaches, the associated confidence intervals exhibit smaller lengths while maintaining adequate coverage rates. The utilities of the proposed method are demonstrated by analyzing a real data set.