A Rasch model analysis of poster-based learning on network flow problem to students’ visual communication ability

Abstract

The network flow problem is one of the subject matters of graph theory application. Given the project assignments to solve real problems network flow problems. So that students have created in terms of visual communication poster-based learning is used to present problem-solving projects. A poster is a media in the form of a simplified illustration of object images that further emphasizes the power of the message, visual, with a combination of colors to express the idea of a particular fact or event. A Rasch measurement model with Winsteps software was used to analyze 54 respondents (male 8, Female 46). Visual communication rubric uses posters with 5 criteria with Rasch analysis results for polytomous data obtained by Person Reliability 0.83. The suitability level of the ideal model of measurement (item fit) is obtained outfit means square (0.5 <MNSQ <1.5, infit means square (0.5 <MNSQ <1.5), and point measure correlation (0.4 <Pt Measure Corr <0.85. To see the map of the distribution of students’ abilities used Wright map (Person-item Map) analysis, while to analyze the students’ abilities the person measure is used which can be seen in the measure order table which sequentially shows the students’ abilities from the highest to the lowest. The DIF items are used to analyze whether there is a bias in gender, in this case, it is obtained that the four criteria of probability values of more than 5% (0.05) there is one whose probability value <0.05 can be explained also by person DIF Plot there are male students who in the fourth criterion the ability is lower than average. The Rasch modeling application can be used to analyze education assessments that are more complex and detailed. The Rasch model can be used to analyze multiple-choice test questions, true/false questions (dichotomous data), essay questions with the same score (polytomous data), and essay questions with different scores for each question (polytomous data with Partial Credit Model).