O objetivo deste trabalho é apresentar os conceitos relacionados a Teste Adaptativo Informatizado, ou abreviadamente TAI, para o modelo logístico unidimensional da Teoria de Resposta ao Item. Utilizamos a abordagem bayesiana para a estimação do parâmetro de interesse, chamado de traço latente ou habilidade. Apresentamos os principais algoritmos de seleção de itens em TAI e realizamos estudos de simulação para comparar o desempenho deles. Para comparação, usamos aproximações numéricas para o Erro Quadrático Médio e para o Vício e também calculamos o tempo médio para o TAI selecionar um item. Além disso, apresentamos como instalar e usar a implementação de TAI desenvolvida neste projeto chamada de TAI2U, que foi desenvolvido no VBA-Excel usando uma interface com o R / The main of this work is to introduce the subjects related to Computerized Adaptive Testing, or breafly CAT, for the unidimensional three-parameter logistic model of Item Response Theory. We use bayesian approach to estimate the parameter of interest. We present several item selection algorithms and we perform simulations comparing them. The comparisons are made in terms of the mean square error, bias of the trait estimates, the average time for item selection and the average length of test. Furthermore, we show how to install e use the CAT implementation of this work called built in MIcrosoft Excel - VBA using interface with the statistical package R

Thesis advisor: Larry H. Ludlow / Item Response Theory (IRT) is a contemporary measurement technique which has been used widely to model testing data and survey data. To apply IRT models, several assumptions have to be satisfied. Local item independence is a key assumption directly related to the estimation process. Many studies have been conducted to examine the impact of local item dependence (LID) on test statistics and parameter estimates in large-scale assessments. However, in the heath care field where IRT is experiencing greater popularity, few studies have been conducted to study LID specifically. LID in the health care field bears some unique characteristics which deserve separate analysis. In health care surveys, it is common to see several items that are phrased in a similar structure or items that have a hierarchical order of difficulties. Therefore, a Guttman scaling pattern, or a deterministic response pattern, is observed among those items. The purposes of this study are to detect whether the Guttman scaling pattern among a subset of items exhibit local dependence, whether such dependence has any impact on test statistics, and whether these effects differ when different IRT models are employed. The score-based approach - forming locally dependent dichotomous items into a polytomous testlet - is used to accommodate LID. Results from this dissertation suggest that the Guttman scaling pattern among a subset of items does exhibit moderate- to high-degree of LID. However, the impact of this special LID is minimal on internal reliability estimates and on the unidimensional data structure. Regardless of which models are employed, the dichotomously-scored person ability estimates are highly correlated with the polytomously-scored person ability estimates. However, the impact of this special LID on test information differs between Rasch models and non-Rasch models. Specifically, when only Rasch models are involved, test information derived from the LID-laden data is underestimated for non-extreme scores; whereas, when non-Rasch models are used, the opposite finding is reached –that is, LID tends to overestimate test information. / Thesis (PhD) — Boston College, 2008. / Submitted to: Boston College. Lynch School of Education. / Discipline: Educational Research, Measurement, and Evaluation.

As pesquisas científicas na área da Psiquiatria freqüentemente avaliam características subjetivas de indivíduos como, por exemplo, depressão, ansiedade e fobias. Os dados são coletados através de questionários, cujos itens tentam identificar a presença ou ausência de certos sintomas associados à morbidade psiquiátrica de interesse. Alguns desses itens, entretanto, podem provocar constrangimento em parte dos indivíduos respondedores por abordarem características ou comportamentos socialmente questionáveis ou, até, ilegais. Um modelo da teoria de resposta ao item é proposto neste trabalho visando diferenciar a relação entre a probabilidade de presença do sintoma e a gravidade da morbidade de indivíduos constrangidos e não constrangidos. Itens que necessitam dessa diferenciação são chamados \\textbf{itens com comportamento diferencial}. Adicionalmente, o modelo permite assumir que indivíduos constrangidos em responder um item possam vir a mentir em suas respostas, no sentido de omitir a presença de um sintoma. Aplicações do modelo proposto a dados simulados para questionários com 20 itens mostraram que as estimativas dos parâmetros são próximas aos seus verdadeiros valores. A qualidade das estimativas piora com a diminuição da amostra de indivíduos, com o aumento do número de itens com comportamento diferencial e, principalmente, com o aumento do número de itens com comportamento diferencial suscetíveis à mentira. A aplicação do modelo a um conjunto de dados reais, coletados para avaliar depressão em adolescentes, ilustra a diferença do padrão de resposta do item ``crises de choro\" entre homens e mulheres. / Psychiatric scientific research often evaluate subjective characteristics of the individual such as depression, anxiety and phobias. Data are collected through questionnaires with items that try to identify the presence or absence of certain symptoms associated with the psychiatric disease. Some of these items though could make some people embarrassed since they are related to questionable or even illegal social behaviors. The item response theory model proposed within this work envisions to differentiate the relationship between the probability of the symptom presence and the gravity of the disease of embarrassed and non-embarrassed individuals. Items that need this differentiation are called differential item functioning (dif). Additionally, the model has the assumption that individuals embarrassed with one particular item could lie across other answers to omit a possible condition. Applications of the proposed model to simulated data for a 20-item questionnaire have showed that parameter estimates of the proposed model are close to their real values. The estimate accuracy gets worse as the number of individuals decreases, the number of dif increases, and especially as the number of dif susceptible to lying increases. The application of the model to a group of real data, collected to evaluate teenager depression, shows the difference in the probability of \"crying crisis\" presence between men and women.

Chi-Keung Leung. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2001. / Includes bibliographical references (p. 138-146). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Mode of access: World Wide Web. / Abstracts in English and Chinese.

An efficient pool is critical for CAT administrations. Two approaches have been developed to design an optimal CAT pool: the linear programming method (LP; Veldkamp & van der Linden, 2000, 2010) and the bin-and-union method (BU; Reckase, 2003, 2010). This study manipulated different content balancing approaches and exposure conditions to investigate their impacts on the pool performances of the LP and BU methods under practical testing situations. The optimal pools were constructed in terms of the specification of an operational fixed-length CAT program and the IRT model employed. This study considered the one-parameter logistic (1PL) model to simulate adaptive test item responses using optimal and operational pools. Several psychometric properties were compared between the pools designed under the LP and BU methods. This research attempted to answer the following question: Under the consideration of content balancing and exposure control, what were the benefits and limitations of the LP and BU methods with respect to the optimal pool design? The results were evaluated in terms of pool characteristics, content constraint management, item exposure control, pool utilization, test reliability, and measurement precision. Similar pool characteristics were found between the LP and BU methods. With respect to the evaluation criteria, the LP and BU pools exhibited consistent performance. However, compared to the LP pools, the BU pools demonstrated slight superiority under the condition with strict content balancing and exposure control. Given two bin widths (.35 and .70), the pools with a bin-width of .35 exhibited better performance than those with a bin-width of .70 with respect to various evaluation criteria. Especially under the condition with the strict content balancing and exposure control, a bin-width of .35 might be a better option to generate an optimal pool than a bin-width of .70 in order to maintain a higher test on-target rate.

For equating tests that measure several distinct proficiencies, procedures that reflect the multidimensional structure of the data are needed. Although there exist a few equating procedures developed under the multidimensional item response theory (MIRT) framework, there is a need for further research in this area. Therefore, the primary objectives of this dissertation are to consolidate and expand MIRT observed score equating research with a specific focus on the common-item nonequivalent groups (CINEG) design, which requires scale linking. Content areas and item types are two focal points of dimensionality. This dissertation uses two studies with different data types and comparison criteria to address the research objectives. In general, a comparison between unidimensional item response theory (UIRT) and MIRT methods suggested a better performance of the MIRT methods over UIRT. The simple structure (SS) and full MIRT methods showed more accurate equating results compared to UIRT. In terms of calibration methods, concurrent calibration outperformed separate calibration for all equating methods under most of the studied conditions.

Thesis (Ph.D.)--University of Delaware, 2006. / Principal faculty advisor: Ratna Nandakumar, School of Education. Includes bibliographical references.

The aim of this study is to investigate the effects of learning styles and cooperative learning on academic achievement of technical university students in calculus. Meanwhile, in order to understand the effectiveness of learning on calculus, we perform item analysis of calculus test conducted in terms of item difficulty and item discrimination analysis. This research is conducted on a quasi-experimental design. Participants in this study are selected from the first year students of two departments in a technical university. Two intact classes of the same department are randomly assigned to be an experimental group and a control group. The experimental group receives the cooperative learning while the other receives the traditional learning. The results show that the learning achievement of active learners is better than reflective learners in the experimental group of information management. On the contrary, the reflective learners have better learning achievement than active learners in the experimental group of computer science and information engineering. The learning achievement of the experimental group is better than the control group in department of information management. The result is opposite in department of computer science and information engineering. But the differences of the learning achievement mentioned above are not statistically significant. Furthermore, item difficulty analysis suggests teachers and students should pay more attentions to those more difficult items and the corresponding calculus sections.

The most recent development in the field of Item Response Theory (IRT) has been the evaluation of IRT models as multilevel models, known as Multilevel IRT models (MLIRT). These models offer several statistical and practical advantages over ordinary IRT models. However, models such as 2-PL MLIRT models have not been studied yet. This dissertation consists of two studies, a simulation and a substantiation for an urban school district dataset. The simulation study tested the performance of twoparameter (2-PL) MLIRT models with predictor variables under various conditions that included 3 test lengths (15, 30, and 60 items), 4 sample sizes (200, 500, 1000, and 2000), 2 correlation conditions between the predictor variable and the ability (or attitude) parameter (rpb=.35 and .8), and 4 binomial distributions of the predictor variable (p=0.1, 0.25, 0.4, and 0.5). The bias and Root Mean Square Deviation (RMSD) values of the item parameters indicated that the distribution of the predictor variable and the correlation between the predictor and the ability (or attitude) parameter did not affect the estimates of 2-PL MLIRT models. These models performed well for sample sizes as low as 500 and test lengths as low as 15 which is lower than the required sample size for ordinary IRT models. Even for a sample size of 200, sufficiently accurate estimates were obtained with more than 300 iterations. The second study investigated the characteristics of the items that measured urban teachers’ perceptions of cultural awareness and beliefs about teaching African American children and tested whether these perceptions were influenced by the teachers’ gender, ethnicity, or teaching experience. Teacher beliefs about teaching African American students, culturally responsive management, and cultural awareness factors were influenced by the ethnicity of the teachers. Culturally responsive management, home and community support, and curriculum and instructional strategies factors were influenced by the teaching experience of the teachers. Items that were biased based on ethnicity or teaching experience were identified. None of the items exhibited gender bias. The study identified items that could be used over other items when the need for a shorter instrument or more informative categories arises.

Computerized adaptive testing attempts to provide the most suitable question for an examinee depending on the examinee¡¦s ability to achieve the best result. Although Maximum Likelihood Estimation (MLE) and Bayesian Likelihood Estimation (BLE) have been provided to solve ability estimation and have good results in the literature, little attention has been paid to the situation when the answer of an item does not conform with the examinee¡¦s ability as expected nor standard derivation changes of the ability estimation. We hypothesized that the Adaptive-Network-Based Fuzzy Inference System (ANFIS) can be used to infer flexible examinee¡¦s ability estimation automically by analyzing the relevant data of the examinee in a test. Consequently, the study presents a novel learning ability model based on ANFIS, which can adaptively choose questions by Item Response Theory. Taking the item discrimination, difficulty, guessing, and the examinee¡¦s ability before he/she answers a question as parameters, the proposed method can infer the adjustment of the examinee¡¦s ability to update its value after he/she answers the question. The ANFIS model of the experiments were developed using MATLAB. The examinees were simulated and the training data were collected under three different situations. Through different combination of ANFIS fuzzy rules, the adjustment of ability is inferred to improve the accuracy of the estimated ability. The error between the true ability and the estimated ability obtained by the proposed model is compared with MLE and BLE. The simulation results show that the estimated ability error of ANFIS is smaller than MLE and BLE when the value of the test information is larger. The proposed method could provide better accuracy of the examinee¡¦s ability and offer more appropriate questions for examinees. Keywords: ANFIS, Item Response Theory, Computerized Adaptive Testing