The Cognitive Diagnostic Assessment Of Junior High School Students’ Reading Comprehension Ability Of Modern Chinese Prose

Reading ability is not only the most basic skill for a person to adapt to future life, but also related to the development of the country and the progress of society. With both the countries and organizations around the world attaching great importance to monitoring and assessing of student’s reading ability, China also has begun to make pilot monitoring of students’ Chinese reading comprehensive ability in primary and middle school since 2011. In order to enhance the reading ability of junior high school students, it is necessary to make process evaluation of their reading ability and then learn about their strengths and weaknesses in reading ability, and thus we can provide students with timely and targeted help. Compared with traditional test which only provides general ability score, cognitive diagnostic assessment can explore the internal psychological process and the skills and strategies behind the score. It can diagnose the strengths and weaknesses of individual cognition so as to help people know about the rules and the mechanisms of mental activity in the learning activities. What’s more, it will help students, teachers, education administrators and parents have a comprehensive understanding of the level of student’s mastery of the knowledge and skills so that they can take remedial measures specifically.The research object of this paper was junior high school students who are in the key stage of the promotion of reading ability. The evaluation target was reading comprehension ability of modern Chinese prose. According to the research paradigm of Cognitive Diagnostic Assessment, this paper adopted Literature Analysis、Expert Evaluation and Think Aloud to make a preliminary definition of cognitive model reflecting the process of reading comprehension of modern Chinese prose of junior high school students. Secondly, based on Q matrix theory, we compiled the cognitive diagnostic test upon cognitive model and then also made further revision about cognitive model. Thirdly, we made cognitive diagnostic assessment for 1301 junior high school students from 27 classes in four schools by the Fusion Model.Finally, we choosed two classes of one school to have a quasi-experimental research, and interviewed the principal, teachers and some students and also have questionnaire survey for them. The outcomes were as follows:(1) The cognitive model of reading comprehension ability of modern Chinese prose of junior high school students was scientific and proper. This cognitive model was composed of seven cognitive attributes which were word decoding, formal schema, information extraction, content analysis, contents generalization, information deduce, and text appreciation. The Attribute hierarchy was unstructured. This model was consistent with psychological process of reading, Chinese Curriculum Standard, test syllabus of Chinese and so on. In addition, it was quite consistent with the result of oral report and fit the observation data very well. Hierarchical Regression Analysis showed that seven cognitive attributes could explain 91.9 percent of variation of item difficulty. The HCI index associated with the cognitive attribute hierarchy was 0.6021. Obviously, the cognitive model was Supported by dual evidences from theory and testing data. The above results showed that the traditional method combined with model revised on the measured data by the Fusion Model was scientific and practical.(2) The cognitive diagnostic test was good statistically. The results of item difficulty analysis by CTT, TRT and CDA consistently showed that difficulty of the test was moderate and the distribution of difficulty was rational(About 30 percent of the items was easy, about 63 percent of the items was moderate and about 7 percent of the items was difficult). The item discrimination analysis by CTT and TRT indicated that 86.67 percent of the items have good distinction ability. The attribute discrimination parameter from CDA showed that 95.89 percent of parameters was less than 0.9.The analysis results by Fusion Model showed that the definition of these attributes of 23 items among 30 items were complete, with the others also complete basically. The reliability coefficient of CTT and TRT were 0.73 and 0.78 respectively. The content validity was good and the theoretic construct validity was 1. Confirmatory Factor Analysis showed that the relationship model between item and property was equipped with ideal fit index(RMSEA was 0.02, GFI and AGFI were 0.97, NFI and NNFI were 0.92 and 0.97 respectively.) and it showed that the construct validity of test was very good.(3) The diagnostic result analysized by Fusion Model was reliable and valid. Firstly, MCMC convergence has occurred to the item parameter and the participant parameters. Secondly, The correlation coefficient between model-predicted values with the observed values of item proportion-correct scores and examinee raw total score were 0.999 and 0.91 respectively and the empirical validation showed that the model fit the data reasonably well. Thirdly, the average Marginal Match Ratio was 0.867, coefficient of internal consistency was 0.671, test-retest reliability was 0.728 and so the diagnostic result was accuracy and dependable. Fourthly, the difference(0.5032) of item proportion-correct scores between masters and non-master declared the good internal validity. The correlation between the posterior probability of mastery(ppm) with the students’ comprehensive performance, the usual language achievement and self-evaluation ability of reading was significant and it showed that the external validity was good.(4) The average probability of seven cognitive properties in reading comprehension of modern Chinese prose were as follows: word decoding(0.8273), formal schema(0.6162), content analysis(0.5185), information extraction(0.4899), contents generalization(0.4768), information deduce(0.4386), text appreciation(0.3923). When making a MANOVA for the group differences on those seven properties of probability, we knowed that the ppm was different among regions, schools, classes, students’ ability level, background and sex. For example, girls outperformed boys, the key class outperformed the regular class, student cadres outperformed the non-student cadres. Except for the A1 attributes, the differences were significant between the other attributes. For example, the key school outperformed the ordinary school, the only child outperformed the non-only child, grade three outperformed grade two, grade two outperformed grade one. However, the difference between grade two and grade three was not significant in content analysis. As for the four surveyed schools, there were differences on five attributes except for word decoding and information deduce. Specifically, performance of the best school was the Railway middle school in Nan Chang, and the second was No 2 middle school in Shangrao, Huangmao middle school and No 2 middle school in Wanzai almost have no differences. As far as the students’ comprehensive performance,the better comprehensive performance, the higher the master degree of five attributes except that there were small differences in word decoding and content generalization. As to the regional differences, we found that urban students were better than the town and rural students in the degree of mastering attributes.(5) The posterior probability of mastery was the result of combined action of individual, family and school. The result of Hierarchical Linear Model analysis indicated that the main influencing factors were students’ comprehensive learning ability, gender and their interests in reading. However, whether or not it is the key class and the grade were the main factors of class.(6) Diagnostic feedback benefited the improvement of students’ reading ability. Compared with the controlled class, the quasi experimental class made a remarkable improvement in the reading skill. The interview with the principal and teachers showed that they agreed with this opinion. The survey’s result about the feedback indicated that the five grades(the degree of mastery were identified by ABCDE) drawed the attention of students. This method could provide richer and more accurate information than two categories(the state of mastery were identified by 0 or 1) and it was consistent with the method which assesses student’s achievement in China at the present. Therefore, it was a scientific, effective, feasible method of reporting diagnostic information.