ABSTRACT
O Exame Nacional do Ensino Médio (Enem) gera uma pontuação para cada domínio que avalia: matemática, linguagens, ciências da natureza e ciências humanas. Reconhecendo a relevância do exame no acesso ao ensino superior e em outros aspectos da vida prática do estudante brasileiro, o presente estudo investiga a fidedignidade dos escores do Enem nos seus quatro domínios. Utilizou-se como amostra os escores dos estudantes que participaram da edição de 2011 do exame. As análises envolveram a estimação dos parâmetros de um modelo de quatro fatores correlacionados e de um modelo bifatorial por meio de análise fatorial confirmatória, além da estimação da fidedignidade composta e da fidedignidade omega dos quatro domínios e do fator geral de desempenho, no caso do modelo bifatorial. Utilizou-se como variáveis observáveis as 30 competências de cada domínio. Os resultados indicaram alta fidedignidade apenas para os escores provenientes do fator geral
The National Exam of Upper Secondary Education (ENEM) generates a score for each domains it assess: mathematics, languages, natural sciences, and humanities. Considering the relevance of the Exam in the access to higher education and in other practical aspects of Brazilian students' life, the present study investigates the reliability of the scores from the four domains. We used as a sample the scores of the students who participated in the 2011 edition of the Exam. The analyzes involved the estimation of the parameters of an oblique four-factor model and a bifactor model using confirmatory factor analysis, as well as the estimation of composite and omega reliability of the four domains and the general performance factor, in the case of the bifactor model. We used the 30 competences of each domain as observable variables. The results indicated high reliability only for the scores from the general factor
El Examen Nacional de la Secundaria Superior (ENEM) genera un puntaje para cada dominio que evalúa: matemática, lenguage, ciencias naturales y humanidades. Puesta la relevancia del Examen en el acceso a la educación superior y en otros aspectos prácticos de la vida de los estudiantes brasileños, el presente estudio investiga la fiabilidad de los puntajes de los cuatro dominios. Usamos como muestra los puntajes de los estudiantes que participaron en la edición 2011 del Examen. Los análisis incluyeron la estimación de los parámetros de un modelo oblicuo de cuatro factores y un modelo bifactorial utilizando análisis factorial confirmatorio, así como la estimación de la confiabilidad compuesta y omega de los cuatro dominios y el factor de desempeño general, en el caso del modelo bifactor. Usamos las 30 competencias de cada dominio como variables observables. Los resultados indicaron una alta fiabilidad solo para los puntajes del factor general
Subject(s)
Students/psychology , Education, Primary and Secondary , PsychometricsABSTRACT
Resumen En el contexto del modelamiento de ecuaciones estructurales se manejan los conceptos de validez convergente y confiabilidad compuesta aplicados a los modelos de medida con ítems congenéricos. Este estudio metodológico tiene como objetivo revisar los puntos de corte estipulados para la Varianza Media Extraída (VME) cuando se usa para establecer la validez convergente de un modelo de medida. Por una parte, se consideró la gran semejanza entre los conceptos de validez convergente y confiabilidad por consistencia interna, al usarse los pesos de medida estandarizados para su estimación. Por otra parte, se analizó la relación entre el número de ítems del factor, la VME y los coeficientes omega y H. A efectos de simplificación, se manejó un peso de medida homogéneo en las demostraciones. Se observó un efecto muy grande del número de ítems, provocando un deterioro en la VME, aun manteniendo constante el mismo nivel de confiabilidad (coeficientes omega y H), y más acusado a mayor nivel de confiabilidad. Se concluye que valores de VME < .50 pueden reflejar niveles aceptables de validez convergente, dependiendo del número de ítems, si incluyen como criterios complementarios: pesos de medida estandarizados ≥ .50 y coeficientes omega y H ≥ .70.
Abstract In the context of the structural equation modeling, the concepts of convergent validity and composite reliability applied to the measurement models with congeneric items are used. This methodological study aims to review the stipulated cut-off points for the Extracted Average Variance (AVE) when it is used to establish the convergent validity of a measurement model. On the one hand, the great similarity between the concepts of convergent validity and internal consistency reliability, when standardized measurement weights are used for their estimation, was considered. On the other hand, the relationships among the number of factor items, the AVE and the omega and H coefficients were analyzed. For simplification purposes, a homogeneous measurement weight was used in the demonstrations. A very large effect of the number of items was observed, causing a deterioration in the AVE, while still maintaining the same level of reliability (omega and H coefficients), and more pronounced at a higher level of reliability. It is concluded that AVE values < .50 may reflect acceptable levels of convergent validity depending on the number of items, if the following criteria are included as complementary criteria: standardized measurement weights ≥ .50, and coefficients omega and H ≥ .70.
Subject(s)
Reproducibility of Results , Analysis of Variance , Review , Latent Class Analysis , Outcome Assessment, Health Care/statistics & numerical data , Latent Class Analysis , MethodsABSTRACT
@#Background: Self-efficacy (SE) is a person’s belief in his or her own capability to perform and accomplish a task that could produce a favourable outcome, despite facing obstacles. This study aimed to confirm the validity and reliability of an SE scale among undergraduate students at the Health Campus of the Universiti Sains Malaysia. Methods: A cross-sectional study was conducted among the undergraduate students using a self-administered questionnaire. After using a purposive sampling method, 562 students completed the questionnaire. Mplus 8 was employed to conduct the confirmatory factor analysis on the psychometric properties of Bandura’s 18-item SE scale with three factors (internal feeling, competing demands and situational). Then, the composite reliability was calculated for each factor. Results: Most of the students were Malay (73.3%) females (79.0%) who exercised 2.62 times a week for an average of 43.37 min per session. The final measurement model was obtained after removing six problematic items, and the model was deemed fit based on several indices [Root Mean Square Error of Approximation (RMSEA) = 0.067, Standardised Root Mean Square Residual (SRMR) = 0.004, Comparative Fit Index (CFI) = 0.924]. The composite reliability values of the three factors were acceptable (0.65 to 0.84). Conclusion: The simplified 12-item SE scale with three factors displayed good fit indices with regard to the data, and they were considered to be acceptable for the current sampl
ABSTRACT
RESUMO A variância média extraída (VME) e a confiabilidade composta (CC) são indicadores associados à qualidade de uma medida. No entanto, é necessário compreender adequadamente a dinâmica dos cálculos da VME e da CC, bem como as suas relações com os conceitos de validade e precisão para evitar equívocos na interpretação dos seus resultados. No presente artigo, ilustramos, por meio de modelos unifatoriais simulados, como o número de itens e a homogeneidade das cargas fatoriais impactam os valores da VME e da CC. Assim, problematizamos a utilização de pontos de cortes fixos para esses indicadores. Além disso, apresentamos argumentos endossando a VME como uma medida de precisão, e não de validade convergente.
ABSTRACT The average variance extracted (AVE) and the composite reliability coefficients (CR) are related to the quality of a measure. Meanwhile, in order to avoid misconceptions, it is required to properly comprehend the equations of the AVE and CR, as well as their relation with the definition of validity and reliability. In this paper, we illustrate, using simulated one-factor models, how the number of items and the homogeneity of factor loadings might influence the AVE and CR results. In so doing, we show that the use of fixed AVE and CR cutoffs is problematic. Moreover, we present reasons endorsing the use of the AVE as a reliability coefficient, instead of a convergent validity index.
ABSTRACT
.
.
.
ABSTRACT
Objective To revise the trait self-control scale for Chinese college students.Methods According to the theory,12 items were selected,then data were collected from a sample of 549 and 133 college students of Harbin.Explorer Factor Analysis,Confirmatory Factor Analysis,reliability test and validity test were used.Results ①Trait self-control scale for college students included eight items,two dimensions were restrained and impulsivity,two dimensions can explain 42.90％ of the total variance.②Confirmatory Factor Analysis showed that the two-dimensional structure was reasonable (x2/df =1.24,CFI =0.97,GFI =0.96,AGFI =0.92,IFI =0.97,TLI =0.96,RMSEA =0.04).③The restrained factor scores were positively correlated with satisfaction of life scores and adaption scores(r=0.23,0.48,P＜0.01),while negatively correlated with negative affect (r=-0.19,P ＜0.01).The impulsivity factor scores were negatively correlated with satisfaction of life scores and adaption scores (r=-0.18,-0.34,P＜0.01),while positively correlated with negative affect(r=0.22,P＜0.01).④The composite reliability coefficients of the scales were good:the coefficient of the subscale restraint was 0.69,and the coefficient of subscale of impulsivity was 0.64,the coefficient of the whole scale was 0.80.The two months test-retest reliability coefficient of 133 subjects was 0.66.Conclusion The revised scale of trait self-control scale has good psychometric quality and can be used in Chinese college students.