ÖZBu çalışmanın amacı açımlayıcı faktör analizinde kullanılan faktör çıkartma yöntemlerini açıklanan varyans ve çıkartılan faktör sayısı bakımından çeşitli örneklem büyüklüğü ve ortak varyans düzeyi koşullarına göre karşılaştırmaktır. Bu çalışma bir simülasyon (benzetim) çalışmasıdır. Bu bağlamda çalışmada kullanılacak veriler SAS programı kullanılarak örneklem büyüklüğüne (100, 500), ortak varyans düzeylerine (yüksek [0.6-0.8], geniş [0.2-0.8], düşük [0.2-0.4]), sabit faktör sayısına (3), sabit değişken sayısına (20) göre 1000 replikasyonla korelasyon matrisleri biçiminde üretilmiştir. Yapılan karşılaştırmalar sonucunda ortak varyansı yüksek veya düşük olarak üretilen veri için örneklem büyüklüğü arttıkça, açıklanan varyans miktarı da artmaktadır. Fakat ortak varyansı geniş olarak üretilen veri grubu için örneklem büyüklüğü arttıkça açıklanan toplam varyansın azaldığı görülmüştür. Analiz edilen tüm durumlarda toplam varyansı en yüksek açıklayan yöntemin temel bileşen analizi ve en düşük açıklayan yöntemin ise imaj faktör yöntemi olduğu gözlenmiştir. Faktör yükleri bakımından yapılan karşılaştırmalara göre temel bileşen analizi her durum için en yüksek faktör yükünü çıkartmıştır. Öte yandan imaj faktör yöntemi ise en düşük faktör yükünü çıkartmıştır. Ortak varyansın geniş aralıkta ya da yüksek aralıkta üretildiği veri için yöntemler arasında faktör sayıları bakımından bir farklılık yoktur. Ortak varyansın düşük olduğu ve örneklem büyüklüğünün küçük olduğu gruplar için imaj faktör yöntemi istenilen faktör sayısını çıkartamamıştır. Dolayısıyla örneklem sayısının ve ortak varyansın düşük olduğu gruplar için imaj faktör yönteminin kullanılması önerilmemektedir. ABSTRACTThe purpose of this study was to compare factor extraction methods used in exploratory factor analysis in terms of variance and factor number according to sample size and the communality level conditions. In this simulation study, data were generated with respect to two levels of sample sizes (100, 500), three levels of communality (high, wide and low), with 20 variables, 3 factors, and Karaman, Atar & Çobanoğlu Aktan 1174 1000 replications by using SAS program. As a resut of comparisons, it was found that total explained variance increased when sample size increased for data with high or low communality levels. However, total explained variance decreased when sample size increased, for the data with wide communality level. Component analysis yielded the biggest total explained variance compared to other methods, whereas image analysis yielded the lowest explained variance. The rest of the factor extracting methods provided the similar results to each other. Component analysis produced the biggest factor loadings and image analysis yielded the lowest factor loadings and there was no difference concerning factor loadings for the remaining methods. For the data generated with high and wide communality levels, no difference was observed among the extraction methods in terms of number of factors. However, for data generated with low communality level image...

No abstract

In this study, differential item functioning (DIF) detection performances of multiple indicators, multiple causes (MIMIC) and logistic regression (LR) methods for dichotomous data were investigated. Performances of these two methods were compared by calculating the Type I error rates and power for each simulation condition. Conditions covered in the study were: sample size (2000 and 4000 respondents), ability distribution of focal group [N(0, 1) and N(-0.5, 1)], and the percentage of items with DIF (10% and 20%). Ability distributions of the respondents in the reference group [N(0, 1)], ratio of focal group to reference group (1:1), test length (30 items), and variation in difficulty parameters between groups for the items that contain DIF (0.6) were the conditions that were held constant. When the two methods were compared according to their Type I error rates, it was concluded that the change in sample size was more effective for MIMIC method. On the other hand, the change in the percentage of items with DIF was more effective for LR. When the two methods were compared according to their power, the most effective variable for both methods was the sample size.

This Monte Carlo simulation study aimed to investigate confirmatory factor analysis (CFA) estimation methods under different conditions, such as sample size, distribution of indicators, test length, average factor loading, and factor structure. Binary data were generated to compare the performance of maximum likelihood (ML), mean and variance adjusted unweighted least squares (ULSMV), mean and variance adjusted weighted least squares (WLSMV), and Bayesian estimators. As a result of the study, it was revealed that increased average factor loading and sample size had a positive effect on the performance of the estimation methods. According to the research findings, it can be said that the methods are sufficient to estimate average factor loading and interfactor correlations, regardless of the estimation methods, in most of the conditions where the average factor loading is 0.7. In small sample sizes particularly, the interfactor correlation was underestimated for skewed indicator conditions. According to the findings of the study, although there is not the most accurate method in all conditions, it can be recommended to use ULSMV method because it performs adequately in more conditions.

Two-level hierarchical linear modelling was used to analyse data obtained from TIMSS 1999 database to examine the effect of inquiry-based learning on science achievement of eighth-grade students in Turkey. The influence of teachers’ emphasis on scientific reasoning and problem solving and availability of school resources for science instruction on inquiry-based learning and science achievement slopes were also examined. At the student-level, overall plausible values were used as the outcome variable whereas gender, availability of home educational resources, attitude toward science, and inquiry-based learning were predictor variables. At the school level, there were two predictor variables: teachers’ emphasis on scientific reasoning and problem solving and availability of school resources for science instruction. While statistically significant and negative relationship was found between inquiry-based learning and within-school science achievement, there were nonsignificant effects of school-level predictors on inquiry-based learning and science achievement slopes. Key words: HLM, inquiry learning, science achievement, TIMSS.