KAĞIT-KALEM VE BİLGİSAYAR FORMUNDA UYGULANAN TESTLERDE ÖĞRENCİ BAŞARISININ KARŞILAŞTIRILDIĞI ÇALIŞMALARIN META-ANALİZİ

Many studies examining on the comparison of computer - based tests and paper - and - pencil tests are available in the literature. These studies are applied to various fields, variables and workgroups. So it is important to compile and rev iew these studies by using meta - analysis for reaerchers and authorities. In this paper, a survey of the studies on comparison of student performances in paper - based and computer - based tests is presented. By considering these studies, a meta - analysis on th e comparison of students’ achievements for paper - based and computer - based tests is provided. To determine if the studies included into meta - analysis have bias, Funnel Graph and Begg - Mazumdar correlation test are utilized for sensitivity analysis. Influence quantities pertaining the studies are analyzed by using Comprehensive Meta - Analysis software. Results of heterogeneity test are analyzed to determine the best effect model and then it is seen that effect quatities shows normal distribution. Based on that, Fixed Effect Model is selected. Hedge’s g coefficient is used to calculate the common effect. There are two basic restrictions of this study. First, due to the lack of time, it could not be possible to reach all studies available for the comparison of pa per - based and computer - based tests. Second, results of the meta - analysis are not compared with respect to the topic fields. It could be better to compare the results for topic fields as future prospect. According to 35 findings related to the 9 different studies included to the meta - analysis, no reasonable difference is observed for paper - based and computer - based tests for 29 comparisons, and some reasonable differences are observed 6 ones. Common influence quantity obtained from all the studies is found as - 0.09. This necessarily means that the differences between scores obtained from paper - based and computer - based tests corresponds to 0.09 standard deviation, which is not a considerable influence quantity. In other words, according to the meta - analysis results, there is no substantial difference for the students’ performance in paper - based and computer - based tests. As a result, even though it is not inconvenient to use paper - based tests, it should be determined while preparing a computer - based test wheth er student achievement is affected by applied test form via some trial applications . Keywords: Computer - based test, paper - based test, meta - analysis.