Search

Comparing language tests and test preparation materials holds important implications for the latter's validity and reliability. However, not enough studies compare such materials across a wide range of indices. Therefore, this study investigated the text complexity of IELTS academic reading tests (IRT) and IELTS reading practice tests (IRPrT). Fine-grained quantitative analyses were undertaken to delineate measures of lexical, syntactic, and discourse complexity across a corpus of 108 IRT and 108 IRPrT published by Pearson, Macmillan, and Cengage Learning. The results suggest little difference between IRT and IRPrT at the lexical level; however, there were significant differences in some measures of syntactic and discourse level complexity. The findings bear implications for stakeholders including learners as test takers, instructors, material developers, and language testing researchers. We interpret this to mean that while IRPrT materials are lexically conducive to the practice for IRT and can provide a similar experience, the IRPrT do show some differences in the amount of subordination and idea repetition at the discourse level. Therefore, instructors and learners may seek to supplement practice with these structures when preparing for IRT, and the designers of such practice materials should consider aligning these factors in the future.

This research goal to develop a multiple-choice closed-ended test to assessing and evaluate students' digital literacy skills. The sample in this study were students at MTsN 1 Blitar City who were selected using a purposive sampling technique. The test was also validated by experts, namely 2 Doctors of Physics and Science from Yogyakarta State University. The test instrument was developed based on five aspects of digital literacy skills: information, communication, content creation, security and problem-solving. Data have been analyzed descriptively and inferentially using the Rasch version and the assist of Quest software. The results showed that eight multiple-choice closed-ended test instruments were declared valid based on expert validation with an Aiken V value of 1.00. The reliability result is 0.97 with a very high category, and the INFIT MNSQ standard deviation value is 0.86-1.16, so seven items are by the Rasch model. Thus, the seven items in the multiple-choice closed-ended test instrument can be used to assessing and evaluate students' digital literacy skills in learning science.

Alternative conceptions become obstacles in physics. However, it is difficult to find instruments that can identify students' alternative conceptions, especially in gases kinetic theory (DIGaKiT). The purpose of this research was to development of diagnostic instrument of DIGaKiT in identifying students' alternative conceptions by Rasch analysis model. The research method used the defining, designing, developing, and disseminating (4D). The samples are 31 students (12 male students and 19 female students, their ages were typically 16 years old) at one of the senior high schools at Belitung. Rasch analysis was used to identify the validity, reliability, and distribution of students' alternative conceptions. The result is that the level of validity and reliability of the instrument is in a good category. Meanwhile, alternative conceptions of the kinetic theory of gases can be identified in all questions, and the questions with the highest alternative conceptions are questions with code Q11 (77%) and the lowest are questions with codes Q1, Q5, and Q6 (4%). Therefore, teachers must design learning processes that can reduce students' alternative conceptions of the kinetic theory of gases material.

This study aims to develop Nature of Science (NOS) measurement instrument for high school students in the Mathematics and Natural Sciences program using COVID-19 cases that had occurred since the end of 2019. This study also attempts to validate test items with the Rasch model approach. The participants in this study are 194 high school students from the Mathematics and Natural Sciences program. Test participants consist of 132 female students and 62 male students. Their age range is 18-21 years old. NOS measurement instrument is presented in multiple choice form, which consists of 16 questions with narration or readings about COVID-19. Data analysis is conducted by dividing the validity into three types, those are content aspect validity, psychometric aspect validity, and construct validity using Rasch model. Content validity is conducted with an assessment involving three experts related to the test material and measured scientific literacy achievement. Psychometric aspects validity involves three psychometricians associated with the construction of the test. Meanwhile, for construct validity, which refers to the concept of construct validity of Messick. The results shows that there are 8 of 16 items in the NOS measurement that occupies all aspects of validity. Regarding to content of construct validity, there are two items do not appropriate with the model, while with substantive validity 94.4% of the test are consistent with the used Rasch model. Regarding to external aspect of construct validity shows that the test instrument can differentiate test participants into two categories, each of which has high and low NOS abilities.

Local item dependence (LID) refers to the situation where responses to items in a test or questionnaire are influenced by responses to other items in the test. This could be due to shared prompts, item content similarity, and deficiencies in item construction. LID due to a shared prompt is highly probable in cloze tests where items are nested within a passage. The purpose of this research is to examine the occurrence and magnitude of LID in a cloze test. A cloze test was analyzed with the Rasch model and locally dependent items were identified with the residual correlations. Findings showed that three pairs of items were locally dependent. When these items were removed from the analysis, test reliability dropped but item fit and unidimensionality improved. Removing the three locally dependent items did not affect person ability mean and standard deviation, though. The findings are discussed in terms of LID detection and modeling in the context of cloze test and language testing.

The purpose of this study was to examine gender and discipline-based Differential Item Functioning (DIF) and Differential Distractor Functioning (DDF) on the Islamic Azad University English Proficiency Test (IAUEPT). The study evaluated DIF and DDF across genders and disciplines using the Rasch model. To conduct DIF and DDF analysis, the examinees were divided into two groups: Humanities and Social Sciences (HSS) and Non-Humanities and Social Sciences (N-HSS). The results of the DIF analysis showed that four out of 100 items had DIF across gender, and two items had discipline DIF. Additionally, gender DDF analysis identified one item each for Options A, B, and C, and four items for Option D. Similarly, the discipline DDF analysis revealed one item for Option A, three items for Option B, four items for Option C, and three items for Option D. The findings of this study have significant implications for test developers. The identification of potential biases in high-stakes proficiency tests can help ensure fairness and equity for all examinees. Furthermore, identifying gender DIF can shed light on potential gender-based gaps in the curriculum, highlighting areas where male or female learners may be disadvantaged or underrepresented in terms of knowledge or skills.

This study aims to compare item and student parameters of dichotomously scored multidimensional constructs estimated based on unidimensional and multidimensional Item Response Theory (IRT) under different conditions of sample size, interdimensional correlation and number of dimensions. This research, conducted with simulations, is of a basic nature. The standard errors of the item and student parameters estimated according to both models were evaluated by the mean of the error squares. The results of the study indicate that there is no significant difference between the errors of the item parameters obtained from the unidimensional and multidimensional IRT in the cases of two-dimensional data structures and high interdimensional correlations. However, when the data structure are three and five dimensional, the item parameter errors resulting from the unidimensional IRT increase. Standard errors for item parameters decreased as the sample size increased. When the standard errors arising from ability parameters were analysIRTed, it was observed that multidimensional IRT estimated with lower errors for all conditions. As a result of the study, it is established that multidimensional IRT provides more accurate results in the analysis of multidimensional constructs, especially in the estimation of student parameters and in making decisions about students. Therefore, it is recommended that multidimensional models should be used for student ability estimation in national and international large-scale examinations.

Language proficiency diagnosis is essential to extract fine-grained information about the linguistic knowledge states and skill mastery levels of test takers based on their performance on language tests. Different from comprehensive standardized tests, many language learning apps often revolve around word-level questions. Therefore, knowledge concepts and linguistic skills are hard to define, and diagnosis must be well-designed. Traditional approaches are widely applied for modeling knowledge in science or mathematics, where skills or knowledge concepts are easy to associate with each item. However, only a few works focus on defining knowledge concepts and skills using linguistic characteristics for language knowledge proficiency diagnosis. In addressing this, we propose a framework for language proficiency diagnosis based on neural networks. Specifically, we propose a series of methods based on our framework that uses different linguistic features to define skills and knowledge concepts in the context of the language learning task. Experimental results on a real-world second-language learning dataset demonstrate the effectiveness and interpretability of our framework. We also provide empirical evidence with comprehensive experiments and analysis to prove that our knowledge concept and skill definitions are reasonable and critical to the performance of our model.

Bayesian Knowledge Tracing (BKT) is a popular interpretable computational model in the educational mining community that can infer a student's knowledge state and predict future performance based on practice history, enabling tutoring systems to adaptively select exercises to match the student's competency level. Existing BKT implementations do not scale to large datasets and are difficult to extend and improve in terms of prediction accuracy. On the other hand, uninterpretable neural network (NN) student models, such as Deep Knowledge Tracing, enjoy the speed and modeling flexibility of popular computational frameworks (e.g., PyTorch, Tensorflow, etc.), making them easy to develop and extend. To bridge this gap, we develop a collection of BKT recurrent neural network (RNN) cells that are much faster than brute-force implementations and are within an order of magnitude of a fast, fine-tuned but inflexible C++ implementation. We leverage our implementation's modeling flexibility to create two novel extensions of BKT that significantly boost its performance. The first merges item response theory (IRT) and BKT by modeling multidimensional problem difficulties and student abilities without fitting student-specific parameters, allowing the model to easily generalize to new students in a principled way. The second extension discovers the discrete assignment matrix of problems to knowledge components (KCs) via stochastic neural network techniques and supports further guidance via problem input features and an auxiliary loss objective. Both extensions are learned in an end-to-end fashion; that is, problem difficulties, student abilities, and assignments to knowledge components are jointly learned with BKT parameters. In synthetic experiments, the skill discovery model can partially recover the true generating problem-KC assignment matrix while achieving high accuracy, even in some cases where the true KCs are structured unfavorably (interleaving sequences). On a real dataset where problem content is available, the skill discovery model matches BKT with expert-provided skills, despite using fewer KCs. On seven out of eight real-world datasets, our novel extensions achieve prediction performance that is within 0.04 AUC-ROC points of state-of-the-art models. We conclude by showing visualizations of the parameters and inferences to demonstrate the interpretability of our BKT RNN models on a real-life dataset.

Scientific thinking constitutes a vital component of scientific competencies, crucial for citizens to adapt to the evolving societal landscape. To cultivate students' scientific thinking, teachers should possess an adequate professional knowledge foundation, which encompasses pedagogical content knowledge (PCK). Assessing teachers' PCK of scientific thinking facilitates the development of effective curricula tailored to their continuous professional development. Despite its significance, empirical studies on biology teachers' PCK of scientific thinking are notably lacking. Hence, this research aimed to create a reliable and valid tool to evaluate upper-secondary school biology teachers' PCK of scientific thinking. The results showed that the instrument exhibits high reliability and good validity, affirming its efficiency for investigative purposes. A collective of 292 in-service biology teachers from upper-secondary schools participated in this investigation through the completion of an online survey. The results indicated that, overall, as well as for each component, upper-secondary school biology teachers' performance on PCK of scientific thinking fell within the lower to middle range. Specifically, the performance levels of four components: knowledge of students (KSU), knowledge of instructional strategies (KIS), knowledge of curriculum (KC), and knowledge of assessment (KA) declined sequentially.

A general framework of latent trait item response models for continuous responses is given. In contrast to classical test theory (CTT) models, which traditionally distinguish between true scores and error scores, the responses are clearly linked to latent traits. It is shown that CTT models can be derived as special cases, but the model class is much wider. It provides, in particular, appropriate modeling of responses that are restricted in some way, for example, if responses are positive or are restricted to an interval. Restrictions of this sort are easily incorporated in the modeling framework. Restriction to an interval is typically ignored in common models yielding inappropriate models, for example, when modeling Likert-type data. The model also extends common response time models, which can be treated as special cases. The properties of the model class are derived and the role of the total score is investigated, which leads to a modified total score. Several applications illustrate the use of the model including an example, in which covariates that may modify the response are taken into account.

Some previous studies suggest that response times (RTs) on rating scale items can be informative about the content trait, but a more recent study suggests they may also be reflective of response styles. The latter result raises questions about the possible consideration of RTs for content trait estimation, as response styles are generally viewed as nuisance dimensions in the measurement of noncognitive constructs. In this article, we extend previous work exploring the simultaneous relevance of content and response style traits on RTs in self-report rating scale measurement by examining psychometric differences related to fast versus slow item responses. Following a parallel methodology applied with cognitive measures, we provide empirical illustrations of how RTs appear to be simultaneously reflective of both content and response style traits. Our results demonstrate that respondents may exhibit different response behaviors for fast versus slow responses and that both the content trait and response styles are relevant to such heterogeneity. These findings suggest that using RTs as a basis for improving the estimation of noncognitive constructs likely requires simultaneously attending to the effects of response styles.

Thurstonian forced-choice modeling is considered to be a powerful new tool to estimate item and person parameters while simultaneously testing the model fit. This assessment approach is associated with the aim of reducing faking and other response tendencies that plague traditional self-report trait assessments. As a result of major recent methodological developments, the estimation of normative trait scores has become possible in addition to the computation of only ipsative scores. This opened up the important possibility of comparisons between individuals with forced-choice assessment procedures. With item response theory (IRT) methods, a multidimensional forced-choice (MFC) format has also been proposed to estimate individual scores. Customarily, items to assess different traits are presented in blocks, often triplets, in applications of the MFC, which is an efficient form of item presentation but also a simplification of the original models. The present study provides a comprehensive review of the present status of Thurstonian forced-choice models and their variants. Critical features of the current models, especially the block models, are identified and discussed. It is concluded that MFC modeling with item blocks is highly problematic and yields biased results. In particular, the often-recommended presentation of blocks with items that are keyed in different directions of a trait proves to be counterproductive considering the goal to reduce response tendencies. The consequences and implications of the highlighted issues are further discussed.

Marginal maximum likelihood, a common estimation method for item response theory models, is not inherently a Bayesian procedure. However, due to estimation difficulties, Bayesian priors are often applied to the likelihood when estimating 3PL models, especially with small samples. Little focus has been placed on choosing the priors for marginal maximum estimation. In this study, using sample sizes of 1,000 or smaller, not using priors often led to extreme, implausible parameter estimates. Applying prior distributions to the c-parameters alleviated the estimation problems with samples of 500 or more; for the samples of 100, priors on both the a-parameters and c-parameters were needed. Estimates were biased when the mode of the prior did not match the true parameter value, but the degree of the bias did not depend on the strength of the prior unless it was extremely informative. The root mean squared error (RMSE) of the a-parameters and b-parameters did not depend greatly on either the mode or the strength of the prior unless it was extremely informative. The RMSE of the c-parameters, like the bias, depended on the mode of the prior for c.

In various fields, including college admission, medical board certifications, and military recruitment, high-stakes decisions are frequently made based on scores obtained from large-scale assessments. These decisions necessitate precise and reliable scores that enable valid inferences to be drawn about test-takers. However, the ability of such tests to provide reliable, accurate inference on a test-taker's performance could be jeopardized by aberrant test-taking practices, for instance, practicing real items prior to the test. As a result, it is crucial for administrators of such assessments to develop strategies that detect potential aberrant test-takers after data collection. The aim of this study is to explore the implementation of machine learning methods in combination with multimodal data fusion strategies that integrate bio-information technology, such as eye-tracking, and psychometric measures, including response times and item responses, to detect aberrant test-taking behaviors in technology-assisted remote testing settings.

Low-stakes test performance commonly reflects examinee ability and effort. Examinees exhibiting low effort may be identified through rapid guessing behavior throughout an assessment. There has been a plethora of methods proposed to adjust scores once rapid guesses have been identified, but these have been plagued by strong assumptions or the removal of examinees. In this study, we illustrate how an IRTree model can be used to adjust examinee ability for rapid guessing behavior. Our approach is flexible as it does not assume independence between rapid guessing behavior and the trait of interest (e.g., ability) nor does it necessitate the removal of examinees who engage in rapid guessing. In addition, our method uniquely allows for the simultaneous modeling of a disengagement latent trait in addition to the trait of interest. The results indicate the model is quite useful for estimating individual differences among examinees in the disengagement latent trait and in providing more precise measurement of examinee ability relative to models ignoring rapid guesses or accommodating it in different ways. A simulation study reveals that our model results in less biased estimates of the trait of interest for individuals with rapid responses, regardless of sample size and rapid response rate in the sample. We conclude with a discussion of extensions of the model and directions for future research.

This article focuses on rater severity and consistency and their relation to major changes in the rating system in a high-stakes testing context. The study is based on longitudinal data collected from 2009 to 2019 from the second language (L2) Finnish speaking subtest in the National Certificates of Language Proficiency in Finland. We investigated whether rater severity and consistency changed over that period and whether the changes could be explained by major changes in the rating system, such as the change of lead examiner, the modus of rating and training (on-site or remote), and the composition of the rater group. The data consisted of 45 rating sessions with 104 raters and 59,899 examinees and were analysed using the many-facet Rasch model and generalized linear mixed models. The analyses indicated that raters as a group became somewhat more lenient over time. In addition, the results showed that the rater community and its practices, the lead examiners, and the modus of rating and training can influence rating behaviour. Finally, we elaborate on implications for both research and practice.

Though a substantial amount of research exists on imputing missing scores in educational assessments, there is little research on cases where responses or scores to an item are missing for all test takers. In this paper, we tackled the problem of imputing missing scores for tests for which the responses to an item are missing for all test takers. We considered three missing-data imputation methods--the median method, the item response theory (IRT) method, and the two-way method--for imputing scores. We compared the performance of these three imputation methods with respect to their accuracy in estimating scaled scores and test reliability for the aforementioned problem. Real data were used in the comparison. All three methods performed well in imputing scaled scores with negligible imputation error: The IRT method and the median method provided slightly more accurate scaled scores. The two-way method provided the most accurate reliability estimates. Recommendations for practice are provided.

Despite the growing interest in incorporating response time data into item response models, there has been a lack of research investigating how the effect of speed on the probability of a correct response varies across different groups (e.g., experimental conditions) for various items (i.e., differential response time item analysis). Furthermore, previous research has shown a complex relationship between response time and accuracy, necessitating a functional analysis to understand the patterns that manifest from this relationship. In this study, response time data are incorporated into an item response model for two purposes: (a) to examine how individuals' speed within an experimental condition affects their response accuracy on an item, and (b) to detect the differences in individuals' speed between conditions in the presence of within-condition effects. For these two purposes, by-variable smooth functions are employed to model differential and functional response time effects by experimental condition for each item. This model is illustrated using an empirical data set to describe the effect of individuals' speed on their reading comprehension ability in two experimental conditions of reading medium (paper vs. digital) by item. A simulation study showed that the recovery of parameters and by-variable smooth functions of response time was satisfactory, and that the type I error rate and power of the test for the by-variable smooth function of response time were acceptable in conditions similar to the empirical data set. In addition, the proposed method correctly identified the range of response time where between-condition differences in the effect of response time on the probability of a correct response were accurate.

For assessment scales applied to different groups (e.g., students from different states; patients in different countries), multigroup differential item functioning (MG-DIF) needs to be evaluated in order to ensure that respondents with the same trait level but from different groups have equal response probabilities on a particular item. The current study compares two approaches for DIF detection: a multiple-group item response theory (MG-IRT) model and a generalized linear mixed model (GLMM). In the MG-IRT model approach, item parameters are constrained to be equal across groups and DIF is evaluated for each item in each group. In the GLMM, groups are treated as random, and item difficulties are modeled as correlated random effects with a joint multivariate normal distribution. Its nested structure allows the estimation of item difficulty variances and covariances at the group level. We use an excerpt from the PISA 2015 reading domain as an exemplary empirical investigation, and conduct a simulation study to compare the performance of the two approaches. Results from the empirical investigation show that the detection of countries with DIF is similar in both approaches. Results from the simulation study confirm this finding and indicate slight advantages of the MG-IRT model approach.

Research has shown that multiple-indicator multiple-cause (MIMIC) models can result in inflated Type I error rates in detecting differential item functioning (DIF) when the assumption of equal latent variance is violated. This study explains how the violation of the equal variance assumption adversely impacts the detection of nonuniform DIF and how it can be addressed through moderated nonlinear factor analysis (MNLFA) model via Bayesian estimation approach to overcome limitations from the restrictive assumption. The Bayesian MNLFA approach suggested in this study better control Type I errors by freely estimating latent factor variances across different groups. Our experimentation with simulated data demonstrates that the BMNFA models outperform the existing MIMIC models, in terms of Type I error control as well as parameter recovery. The results suggest that the MNLFA models have the potential to be a superior choice to the existing MIMIC models, especially in situations where the assumption of equal latent variance assumption is not likely to hold.

Purpose: The aim of the present study was to compare scale and conditional reliability derived from item response theory analyses among the most commonly used, as well as several newly developed, observation, interview, and parent-report autism instruments. Methods: When available, data sets were combined to facilitate large sample evaluation. Scale reliability (internal consistency, average corrected item-total correlations, and model reliability) and conditional reliability estimates were computed for total scores and for measure subscales. Results: Generally good to excellent scale reliability was observed for total scores for all measures, scale reliability was weaker for RRB subscales of the ADOS and ADI-R, reflecting the relatively small number of items for these measures. For diagnostic measures, conditional reliability tended to be very good (> 0.80) in the regions of the latent trait where ASD and non-ASD developmental disability cases would be differentiated. For parent-report scales, conditional reliability of total scores tended to be excellent (> 0.90) across very wide ranges of autism symptom levels, with a few notable exceptions. Conclusions: These findings support the use of all of the clinical observation, interview, and parent-report autism symptom measures examined, but also suggest specific limitations that warrant consideration when choosing measures for specific clinical or research applications.

A modified and improved inductive inferential approach to evaluate item discriminations in a conditional maximum likelihood and Rasch modeling framework is suggested. The new approach involves the derivation of four hypothesis tests. It implies a linear restriction of the assumed set of probability distributions in the classical approach that represents scenarios of different item discriminations in a straightforward and efficient manner. Its improvement is discussed, compared to classical procedures (tests and information criteria), and illustrated in Monte Carlo experiments as well as real data examples from educational research. The results show an improvement of power of the modified tests of up to 0.3.

In the existing multidimensional extensions of the log-normal response time (LNRT) model, the log response times are decomposed into a linear combination of several latent traits. These models are fully compensatory as low levels on traits can be counterbalanced by high levels on other traits. We propose an alternative multidimensional extension of the LNRT model by assuming that the response times can be decomposed into two response time components. Each response time component is generated by a one-dimensional LNRT model with a different latent trait. As the response time components--but not the traits--are related additively, the model is partially compensatory. In a simulation study, we investigate the recovery of the model's parameters. We also investigate whether the fully and the partially compensatory LNRT model can be distinguished empirically. Findings suggest that parameter recovery is good and that the two models can be distinctly identified under certain conditions. The utility of the model in practice is demonstrated with an empirical application. In the empirical application, the partially compensatory model fits better than the fully compensatory model.

Given the frequent presence of slipping and guessing in item responses, models for the inclusion of their effects are highly important. Unfortunately, the most common model for their inclusion, the four-parameter item response theory model, potentially has severe deficiencies related to its possible unidentifiability. With this issue in mind, the dyad four-parameter normal ogive (Dyad-4PNO) model was developed. This model allows for slipping and guessing effects by including binary augmented variables--each indicated by two items whose probabilities are determined by slipping and guessing parameters--which are subsequently related to a continuous latent trait through a two-parameter model. Furthermore, the Dyad-4PNO assumes uncertainty as to which items are paired on each augmented variable. In this way, the model is inherently exploratory. In the current article, the new model, called the Set-4PNO model, is an extension of the Dyad-4PNO in two ways. First, the new model allows for more than two items per augmented variable. Second, these item sets are assumed to be fixed, that is, the model is confirmatory. This article discusses this extension and introduces a Gibbs sampling algorithm to estimate the model. A Monte Carlo simulation study shows the efficacy of the algorithm at estimating the model parameters. A real data example shows that this extension may be viable in practice, with the data fitting a more general Set-4PNO model (i.e., more than two items per augmented variable) better than the Dyad-4PNO, 2PNO, 3PNO, and 4PNO models.

Evaluating scale comparability in international large-scale assessments depends on measurement invariance (MI). The root mean square deviation (RMSD) is a standard method for establishing MI in several programs, such as the Programme for International Student Assessment and the Programme for the International Assessment of Adult Competencies. Previous research showed that the RMSD was unable to detect departures from MI when the latent trait distribution was far from item difficulty. In this study, we developed three alternative approaches to the original RMSD: equal, item information, and b-norm weighted RMSDs. Specifically, we considered the item-centered normalized weight distributions to compute the item characteristic curve difference in the RMSD procedure more efficiently. We further compared all methods' performance via a simulation study and the item information and b-norm weighted RMSDs showed the most promising results. An empirical example is demonstrated, and implications for researchers are discussed.

The cross-classified data structure is ubiquitous in education, psychology, and health outcome sciences. In these areas, assessment instruments that are made up of multiple items are frequently used to measure latent constructs. The presence of both the cross-classified structure and multivariate categorical outcomes leads to the so-called item-level data with cross-classified structure. An example of such data structure is the routinely collected student evaluation of teaching (SET) data. Motivated by the lack of research on multilevel IRT modeling with crossed random effects and the need of an approach that can properly handle SET data, this study proposed a cross-classified IRT model, which takes into account both the cross-classified data structure and properties of multiple items in an assessment instrument. A new variant of the Metropolis-Hastings Robbins-Monro (MH-RM) algorithm was introduced to address the computational complexities in estimating the proposed model. A preliminary simulation study was conducted to evaluate the performance of the algorithm for fitting the proposed model to data. The results indicated that model parameters were well recovered. The proposed model was also applied to SET data collected at a large public university to answer empirical research questions. Limitations and future research directions were discussed.

Categorical growth models describe examinee growth in terms of performance-level category transitions, which implies that some percentage of examinees will be misclassified. This paper introduces a new procedure for estimating the classification accuracy of categorical growth models, based on Rudner's classification accuracy index for item response theory-based assessments. Results of a simulation study are presented to provide evidence for the accuracy and validity of the approach. Also, an empirical example is presented to demonstrate the approach using data from the Indiana Student Performance Readiness and Observation of Understanding Tool growth model, which classifies examinees into growth categories used by the Office of Special Education Programs to monitor the progress of preschool children who receive special education services.

Identifying items with differential item functioning (DIF) in an assessment is a crucial step for achieving equitable measurement. One critical issue that has not been fully addressed with existing studies is how DIF items can be detected when data are multilevel. In the present study, we introduced a Lord's Wald X[superscript 2] test-based procedure for detecting both uniform and non-uniform DIF with polytomous items in the presence of the ubiquitous multilevel data structure. The proposed approach is a multilevel extension of a two-stage procedure, which identifies anchor items in its first stage and formally evaluates candidate items in the second stage. We applied the Metropolis-Hastings Robbins-Monro (MH-RM) algorithm to estimate multilevel polytomous item response theory (IRT) models and to obtain accurate covariance matrices. To evaluate the performance of the proposed approach, we conducted a preliminary simulation study that considered various conditions to mimic real-world scenarios. The simulation results indicated that the proposed approach has great power for identifying DIF items and well controls the Type I error rate. Limitations and future research directions were also discussed.

A Monte Carlo simulation study was conducted to compare fit indices used for detecting the correct latent class in three dichotomous mixture item response theory (IRT) models. Ten indices were considered: Akaike's information criterion (AIC), the corrected AIC (AICc), Bayesian information criterion (BIC), consistent AIC (CAIC), Draper's information criterion (DIC), sample size adjusted BIC (SABIC), relative entropy, the integrated classification likelihood criterion (ICL-BIC), the adjusted Lo-Mendell-Rubin (LMR), and Vuong-Lo-Mendell-Rubin (VLMR). The accuracy of the fit indices was assessed for correct detection of the number of latent classes for different simulation conditions including sample size (2,500 and 5,000), test length (15, 30, and 45), mixture proportions (equal and unequal), number of latent classes (2, 3, and 4), and latent class separation (no-separation and small separation). Simulation study results indicated that as the number of examinees or number of items increased, correct identification rates also increased for most of the indices. Correct identification rates by the different fit indices, however, decreased as the number of estimated latent classes or parameters (i.e., model complexity) increased. Results were good for BIC, CAIC, DIC, SABIC, ICL-BIC, LMR, and VLMR, and the relative entropy index tended to select correct models most of the time. Consistent with previous studies, AIC and AICc showed poor performance. Most of these indices had limited utility for three-class and four-class mixture 3PL model conditions.

Forced-choice (FC) measures have been widely used in many personality or attitude tests as an alternative to rating scales, which employ comparative rather than absolute judgments. Several response biases, such as social desirability, response styles, and acquiescence bias, can be reduced effectively. Another type of data linked with comparative judgments is response time (RT), which contains potential information concerning respondents' decision-making process. It would be challenging but exciting to combine RT into FC measures better to reveal respondents' behaviors or preferences in personality measurement. Given this situation, this study aims to propose a new item response theory (IRT) model that incorporates RT into FC measures to improve personality assessment. Simulation studies show that the proposed model can effectively improve the estimation accuracy of personality traits with the ancillary information contained in RT. Also, an application on a real data set reveals that the proposed model estimates similar but different parameter values compared with the conventional Thurstonian IRT model. The RT information can explain these differences.

In recent years, some models for binary and graded format responses have been proposed to assess unipolar variables or "quasi-traits." These studies have mainly focused on clinical variables that have traditionally been treated as bipolar traits. In the present study, we have made a proposal for unipolar traits measured with continuous response items. The proposed log-logistic continuous unipolar model (LL-C) is remarkably simple and is more similar to the original binary formulation than the graded extensions, which is an advantage. Furthermore, considering that irrational, extreme, or polarizing beliefs could be another domain of unipolar variables, we have applied this proposal to an empirical example of superstitious beliefs. The results suggest that, in certain cases, the standard linear model can be a good approximation to the LL-C model in terms of parameter estimation and goodness of fit, but not trait estimates and their accuracy. The results also show the importance of considering the unipolar nature of this kind of trait when predicting criterion variables, since the validity results were clearly different.

In an online learning environment, it is important to establish a suitable assessment approach that can be adapted on the fly to accommodate the varying learning paces of students. At the same time, it is essential that assessment criteria remain compliant with the expected learning outcomes of the relevant education standard which predominantly utilizes a competency-based curriculum such as in Indonesia. The aim of the research in this paper is to improve the adaptiveness of questions in the existing Computerized Adaptive Testing (CAT) model by taking into consideration multiple aspects of user context. We propose a context-based question selection model based on competency evaluation by merging four methods of Classical Test Theory, Rasch Model, Linear and Quadratic models, and the combination of branching and item adaptive methods to select questions of suitable difficulty for each individual student. To evaluate the proposed model, we conducted experiments based on a real dataset of 689 elementary school students in Indonesia. The experiment results prove the effectiveness of the proposed model in terms of accuracy in predicting the appropriateness of the questions in relation to the students' ability. This adaptive assessment method which accurately builds upon the students' competency level will support students' success in the online learning environment.

Previous research reveals that the visual design of open-ended questions should match the response task so that respondents can infer the expected response format. Based on a web survey including specific probes in a list-style open-ended question format, we experimentally tested the effects of varying numbers of answer boxes on several indicators of response quality. Our results showed that using multiple small answer boxes instead of one large box had a positive impact on the number and variety of themes mentioned, as well as on the conciseness of responses to specific probes. We found no effect on the relevance of themes and the risk of item non-response. Based on our findings, we recommend using multiple small answer boxes instead of one large box to convey the expected response format and improve response quality in specific probes. This study makes a valuable contribution to the field of web probing, extends the concept of response quality in list-style open-ended questions, and provides a deeper understanding of how visual design features affect cognitive response processes in web surveys.

In order to investigate the influence of separation of grade distributions and ratio of common items on the precision of vertical scaling, this simulation study chooses common item design and first grade as base grade. There are four grades with 1,000 students each to take part in a test which has 100 items. Monte Carlo simulation method is used to simulate the response matrices by self-made program in R 3.0. As the items are scored by 0/1, we select two-parameter logistic model of Item Response Theory and use BILOG-MG for concurrent calibration with EAP method. The Bias and RMSE are calculated as precision indicators. The results show that: (1) Estimation precision of item and ability parameters differs in different grades. For discrimination and difficulty parameters, estimation precision is higher as closer to the base grade and is lower with the increase of effect size. For the ability parameters, the estimation precision is high generally except for fourth grade which is much lower. The precision is best at 0.5 of effect size in general. (2) There is an interaction between the ratio of common items to total test and effect size. When the effect size is 0.5 and 1.0, estimation precision of each grade is most accurate at 30% of common-item ratio. When the effect size is 1.5, the estimation precision of difficulty parameters is best for first, second, and third grade at 30% of common-item ratio while grade 4 at 15% of common-item ratio. The ability parameters of all grades are all best estimated at 15% of common item ratio. There must be a trade-off between the estimation precision of ability parameters and item parameter if the common item ratio is at the range of 15% to 30%. (3) The choice of base grade affects the accuracy of vertical scaling. When the lower grade is selected as the base grade, if the number of consecutive cumulative conversions from the upper grade test score to the lower grade exceeds 2, there will be a large deviation. Therefore if the senior grade changes to the junior grade, it is suggested that the gap of grades should not exceed 2 grades. As a whole, the proportion of anchor items for vertical scaling is set at 30%, but it is better to set the proportion of anchor items as "variable" value (15%-30%) when considering the separation of grade distributions.

The Digital Stress Scale (DSS) is used to measure digital stress, which is the perceived stress and anxiety associated with social media use. In this study, the Chinese version of the DSS was validated using a sample of 721 Chinese college students, 321 males and 400 females (KMO = 0.923; Bartlett = 5,058.492, p < 0.001). Confirmatory factor analysis (CFA) and item response theory (IRT) revealed that the five-factor model fits our data best. And the good criterion validity was shown through the association of the Chinese version of the DSS with anxiety and depression. The results of measurement invariance supported that the scale has measurement equivalence across the region (rural or urban). However, they did not support strict measurement invariance across gender. Overall, our results suggest that the DSS is an ideal tool for measuring digital stress in a Chinese sample. People can benefit from assessing digital stress, which has the potential to moderate the relationship between social media use and mental health. These psychometric investigations of the DSS will allow future research to investigate construct validity better.

Introduction/Background: In 2016, the Adolescent Motor Competence Questionnaire (AMCQ) was developed in Australia. The AMCQ was developed to assess perceived motor competence in adolescents between the ages of 12 and 18 years. The 26 items represent four factors: participation in physical activity and sports, activities of daily living, public performance, and peer comparison. Aims/Objectives: Currently, no validated German-language self-report screening instrument exists for assessing perceived motor competence among adolescents. This study aimed to validate the German version of the AMCQ (AMCQ-GER). Methods/Approach: The German translation's factor structure and psychometric properties were analyzed based on a community sample of 635 healthy children and adolescents between 10 and 18 years of age (12.9 ± 2.31 years, 321 boys and 312 girls) and 100 healthy young adults between 19 and 30 years of age (23.0 ± 2.92 years, 49 men and 51 women). The previously validated cutoff score of [less than or equal to]83 out of 104 was used to group the sample into high and low motor competence. Results: The mean AMCQ-GER score was 84.9 (SD = 8.59), and 38.8% of the participants identified with low motor competence. The principle component analysis revealed clear evidence for a three-factorial structure comprising physical activity and sports, general clumsiness, and activities of daily living (with an explained variance of 29.2%). Reliability for all scales was excellent, with a McDonald's ? of at least 0.70. The overall conclusion of the Rasch analysis supported the confirmatory factor analysis, although with an overall smaller number of items (18 instead of 26). The AMCQ-GER was positively correlated with age (0.26**) and participation in organized club sports (0.19**). It was negatively correlated with body mass index (-0.23**) and school grade in physical activity (-0.40**). Conclusions/Relevance: Our rigorous validation protocol has generated a remarkable reproduction of the AMCQ in German. These results suggest that the language and structure of the questionnaire is appropriate for German-speaking countries.

Item response theory (IRT) refers to a family of mathematical models which describe the relationship between latent continuous variables (attributes or characteristics) and their manifestations (dichotomous/polytomous observed outcomes or responses) with regard to a set of item characteristics. Researchers typically use parametric IRT (PIRT) models to measure educational and psychological latent variables. However, PIRT models are based on a set of strong assumptions that often are not satisfied. For this reason, non-parametric IRT (NIRT) models can be more desirable. An exploratory NIRT approach is kernel smoothing IRT (KS-IRT; Ramsay, 1991) which estimates option characteristic curves by non-parametric kernel smoothing technique. This approach only gives graphical representations of item characteristics in a measure and provides preliminary feedback about the performance of items and measures. Although KS-IRT is not a new approach, its application is far from widespread, and it has limited applications in psychological and educational testing. The purpose of the present paper is to give a reader-friendly introduction to the KS-IRT, and then use the KernSmoothIRT package (Mazza et al., 2014, 2022) in R to straightforwardly demonstrate the application of the approach using data of Children's Test Anxiety scale.

Unidimensionality is an important assumption of measurement but it is violated very often. Most of the time, tests are deliberately constructed to be multidimensional to cover all aspects of the intended construct. In such situations, the application of unidimensional item response theory (IRT) models is not justifieddue to poor model fit and misleading results. Multidimensional IRT (MIRT) models can handle several dimensions simultaneously and yield person ability parameters on several dimensions which is helpful for diagnostic purposes too. Furthermore, MIRT models use the correlation between the dimensions to enhance the precision of the measurement. In this study, a reading comprehension test is modeled with the multidimensional Rasch model. The findings showed that a correlated 2-dimensional model has the best fit to the data. The bifactor model revealed some interesting information about the structure of reading comprehension and the reading curriculum. Implications of the study for the testing and teaching of reading comprehension are discussed.

Reading aloud is recommended as a simple technique to measure speaking ability (Hughes & Hughes, 2020; Madsen, 1983). Reading aloud is currently used in the Pearson Test of English and a couple of other international English as a second language proficiency tests. Due to the simplicity of the technique, it can be used in conjunction with other techniques to measure foreign and second language learners' speaking ability. One issue in reading aloud as a testing technique is its psychometric modeling. Because of the peculiar structure of reading-aloud tasks, analyzing them with item response theory models is not straightforward. In this study, the Rasch partial credit model (PCM) is suggested and used to score examinees' reading-aloud scores. The performances of 196 foreign language learners on five reading-aloud passages were analyzed with the PCM. Findings showed that the data fit the RPCM well and the scores are highly reliable. Implications of the study for psychometric evaluation of reading aloud or oral reading fluency are discussed.

The Multiple-choice (MC) item format is commonly used in educational assessments due to its economy and effectiveness across a variety of content domains. However, numerous studies have examined the quality of MC items in high-stakes and higher-education assessments and found many flawed items, especially in terms of distractors. These faulty items lead to misleading insights about the performance of students and the final decisions. The analysis of distractors is typically conducted in educational assessments with multiple-choice items to ensure high-quality items are used as the basis of inference. Item response theory (IRT) and Rasch models have received little attention for analyzing distractors. For that reason, the purpose of the present study was to apply the Rasch model, to a grammar test to analyze items' distractors of the test. To achieve this, the current study investigated the quality of 10 instructor-written MC grammar items used in an undergraduate final exam, using the items responses of 310 English as a foreign language (EFL) students who had taken part in an advanced grammar course. The results showed an acceptable fit to the Rasch model and high reliability. Malfunctioning distractors were identified.

This study aimed at evaluating a PISA-like reading test developed by teachers participating in the teacher training for teaching PISA-like reading. To serve this purpose, an experimental test was administered to 107 students aged 15-16 using a set of text and questions constructed according to the criteria of the PISA Reading test Level 1. Item analysis was performed following the sampling using Rasch Measurement, deemed essential for determining the ideal index of test items relative to students' ability in making the correct response. The component of the calculation comprises reliability, separation, dan standard error. The Rasch model was constructed manually using Microsoft Excel to obtain the result of the calculation, and a Wright Map was also made manually to illustrate the result of the calculation. The results of the item analysis indicate that the test and the items the teachers constructed have met good criteria. The results revealed an even distribution of test item difficulty at the targeted level. The samples' ability to make correct answers, however, was decentralized towards the test items of moderate level of difficulty. Only a limited number of students showed good ability in their response to test items of higher difficulty. These findings have the practical implication of advancing PISA-like reading test teaching and writing models by providing more information on teachers' ability to write PISA-like reading test items and the levels of difficulty of the items written by the teachers, as indicated by the students' responses to the test items.

Language placement tests (LPTs) are used to assess students' proficiency in a progressive manner in the target language. Based on their performance, students are assigned to stepped language courses. These tests are usually considered low stakes because they do not have significant consequences in students' lives, which is perhaps the reason why studies conducted with LPTs are scarce. Nevertheless, tests should be regularly examined, and statistical analysis should be conducted to assess their functioning, particularly when they have a medium or highstakes impact. In the case of LPTs administered on a large-scale, the logistic and administrative consequences of an ill-defined test may lead to an economic burden and unnecessary use of human resources which can also affect students negatively. This study was undertaken at one of the largest public institutions in Latin America. Nearly 1700 students sit an English LPT every academic semester. A diagnostic statistical analysis revealed a need for revision. To retrofit the test, a new test architecture and blueprints were designed in adherence to the new curriculum created following the Common European Framework of Reference for Languages. After the institution gave two courses to language instructors in language assessment, new items were developed and tried out gradually in several pilot studies conducted with a sample of actual examinees. Then, Item Response Theory (IRT) was used to examine the functioning of the new test items. The aim of this study is to show how the test was retrofitted, and to compare the functioning of the retrofitted version of the English LPT with the previous one. The results show that the quality of items was higher than that of the former English LPT. This study has implications for the design of language tests administered large-scale in higher education, particularly in (semi) periphery countries that decide to design and administer their own LPTs.

Although the Rasch model is used to measure latent traits like attitude or ability where there are multiple latent structures within the dataset it is best to use a technique called the Mixture Rasch Model (MRM) which is a combination of a Rasch model and a latent class analysis (LCA). This study used data from a survey for teachers, teacher candidates, and teacher education program faculty with a sample of 296 candidates, 648 graduates, and 501 program faculty. Survey items based on these competencies were administered to in one Western state to ascertain how well the program candidates attended prepared them for the teaching profession. The 40 items common to surveys of the three groups were submitted to mixture Rasch analysis to determine whether distinct patterns of item response were discernible. Analyses yielded two classes which brings the construct validity of the survey into question. Results showed that the Mixture Rasch Model is and can be useful to determine subgroups for survey researchers. This research presents a demonstration of usefulness of the Mixture Rasch Model for the analysis of survey data.

Statistical process control (SPC) charts have been widely used in the field of educational measurement. The cumulative sum (CUSUM) is an established SPC method to detect aberrant responses for educational assessments. There are many studies that investigated the performance of CUSUM in different test settings. This paper describes the CUSUM procedure and shows how it can be used to monitor the testtaking process and detect aberrant responses. It aims to provide an accessible guide for the CUSUM control chart.