Year : 2018 Volume : 14 Issue : 28

Full Text (PDF)

Lojistik Regresyon Analizi ile Pisa Araştırmasında Öğrenci Başarısının Modellenmesi

Open Access

Abstract

Bu çalışmanın amacı Türkiye örneklemindeki öğrencilerin PISA sınavındaki başarılarını binary lojistik regresyon ile modellemektir. Açıklayıcı değişken olarak bazı sosyo-kültürel özellikler kullanımış olup Bağımlı değişken iki kategorilidir ve öğrenci puanının OECD ortalaması üzerinde olup olmamasını belirtmektedir. Veriler PISA-2009 Türkiye örneklemine aittir. PISA 2009 Türkiye örneklemi, okul türlerine göre tabakalı rastgele yöntemle belirlenen toplam 170 okuldan 4996 öğrenciden oluşmaktadır. Katılımcıların “başarı” grubuna girme ihtimali üzerine cinsiyet, evde konuşulan dil, evde kitap sayısı, bölge, ebeveyn eğitim düzeyi, bilgisayara yönelik tutum, okula yönelik tutum ve varlık indeksinin etkilerini saptamak için lojistik regresyon analizi uygulanmıştır. Model istatistiksel olarak anlamlı bulunmuştur. Modelin açıklanabilen değişkenliği %23.8 (Nagelkerke R2), doğru sınıflandırma oranı %67,9’dir. Analiz sonuçlarına göre kızların başarılı grupta yer alması erkeklere göre 1,71 kat daha olasıdır. Evde konuştuğu dil Türkçe olan öğrencinin başarılı grupta yer alması diğerlerine göre 1,65 kat daha olasıdır. Evde bulunan kitap sayısının görece yüksek olması, ebeveyn eğitim düzeyi, bilgisayar tutumu ve varlık indeksi (WEALTH) başarılı grupta yer alma olasılığının yükselmesi ile ilişkili bulunmuştur. Okula yönelik tutum anlamlı bulunamadı.

Keywords

Lojistik-regresyon   PISA-araştırması   Öğrenci-başarısı  

Corresponding Author

Recep Bindak

Authors

Recep Bindak

Details

DOI 10.26650/ekoist.2018.14.28.0010

Submission : 30 Mar 2017

Early Viewed : 20 Eyl 2018

Full Text (PDF)

Modeling Students’ Achievement in PISA Research with Logistic Regression Analysis

Open Access

Abstract

The aim of this study is to the model the achievements of students in the Turkey sample on the PISA exam with binary logistic regression. Some socio-cultural variables were used as explanatory variables. The dependent variable is dichotomous and indicates whether the student score is above the OECD average. The data were from PISA 2009. The PISA 2009 Turkey sample consists of 4,996 students from 170 schools selected by stratified randomized sampling according to school types. Logistic regression analysis was performed to ascertain the effects of gender, home language, number of books at home, statistical region, parents’ education level, attitude toward computer, attitude toward school, and family wealth index on the likelihood that participants belong to the “success” group. The model was statistically significant. The model explained 23.8% (Nagelkerke R2) of the variance in success and correctly classified 67.9% of cases. According to analysis, girls were 1.71 times more likely to exhibit success than boys. Students whose home language was Turkish were 1.65 times more likely to exhibit success than students with other home languages. Increase in the number of books at home, parents’ educational level, attitude toward computer, and family wealth (WEALTH) index were associated with an increase in the likelihood of belonging to success group. Attitude toward school was not statistically significant.

Abstract Extended

Program for International Student Assessment (PISA) is an international research project organized by the OECD, in which the knowledge and skills of students are evaluated. This research, which is performed in countries covering approximately 90% of the world economy, is carried out every three years. The PISA survey assesses the extent to which 15-year-old students in OECD member and nonmember countries have the basic knowledge and skills needed to be relevant in modern society. In the PISA survey, information pertaining to the reading skills and mathematical and science literacies of the students as well as their views about themselves, their motivation, families, and schools, are collected using measuring tools containing various types of questions. PISA is the most comprehensive and detailed international program that assesses students’ performance and collects the data necessary to explain the differences in students’ performances. The results of the PISA project significantly contribute to understanding both the international and domestic educational outcomes. Therefore, the in-depth examination of the data obtained from the PISA project has become an important research area; this data include records of the obtained results, and it is shared with the scientific community. The degrees of reliability and validity of PISA results are high, and the results significantly aids the understanding of educational outcomes. The purpose of this study is to examine the relationship of students’ achievements in mathematics, reading, and science with psychological, sociological, and cultural variables using the data from PISA 2009 Turkey sample. The data obtained from the questionnaires and tests administered to the students, within the Turkey sample, under the PISA 2009 program are used in this study. This data was obtained from the database at www.pisa2009.acer.edu.au, which is available to all researchers. PISA 2009 Turkey sample consists of 56 provinces from 12 units of statistical regions (NUTS) and 4,996 students from 170 schools, which were selected by stratified random sampling according to the type of school. Method The relationship between students’ achievement in mathematics, reading, and science and the psychological, sociological, and cultural variables was examined by forming logistic regression models. Logistic regression analysis is a regression method that is used to develop classification and assignment processes. Dependent variables: (i) 1 = group (1). At least one of the mathematics or reading or science score is equal to or above the OECD average;(ii) 0 = group (0). Mathematics or reading or science score is below the OECD average. (iii) The independent variables used in the study are as follows: (iv) X1 = Gender (1 = female; reference category = male) (v) X2 = Home language (1= Turkish; reference category = other languages) (vi) X3 = Statistical region (reference category = TRC) (vii) X4 = Number of books at home (viii) X5 = Levels of parents’ education (total in years) (ix) X6 = Attitude toward computer (x) X7 = Attitude toward school (xi) X8 = Wealth index The relationship between the dependent and independent variables was demonstrated by binary logistic regression analysis. Results The omnibus test statistics of the models were significant (c2 = 849.01; p < 0.01). In addition, the models were observed to have a good fit in terms of determination (Hosmer - Lemeshow test statistic was 10.247; df = 8; p = 0.248). Table 8 contains the coefficients, which show the relationship of all the variables in the analysis with dependent variables; the p-values of Wald test statistics, which determine whether standard errors of coefficients and coefficients are statistically significant; and the odds ratios. According to the logistic regression analysis results, the X1, X2, X3, X4, X5, X6, and X8 variables had a significant effect in determining the “success” group. The gender variable (X1) was found to be associated with the success group, while the other variables were held constant. The possibility of the female students to be in the success group was 1.711 times higher than that for the male students. The home language variable (X2) was found to be associated with the success group. The possibility of the students who spoke Turkish at home to be in the success group was 1.654 times higher than that for students who spoke other languages. The statistical region variable (X3) was found to be associated with the success group. Additionally the independent variables such as the number of the books at home (X4), the parents’ education levels (X5), the attitude toward computer (X6), and the wealth index (X8) were associated with the possibility of the student to be in the success group. However, the variable of attitude toward school (X7) did not significantly affect the mentioned dependent variable.A correct classification rate of the models was created with a cut-off value of 0.5. The correct classification rate was obtained as 67.9%. Accordingly, using the independent variables in the model, whichever of the groups (pass or fail) the student will be assigned to in these areas could be estimated with an accuracy of 67.9%. It is clear that PISA is quite an extensive application in terms of both the measured schools, environment, student characteristics, and the sample in which it is applied anywhere in the world. Therefore, it can be said that the output of the program depicts the educational outcomes of the participating countries.

Keywords

Logistic-regression   PISA-survey   Student-achievement  

Corresponding Author

Recep Bindak

Authors

Recep Bindak

References

  • Acar, T. (2012). 2009 yılı uluslararası öğrenci başarılarını değerlendirme programında Türk öğrencilerin başarılarını etkileyen faktörler. Eğitimde ve Psikolojide Ölçme ve Değerlendirme Dergisi, 3(2), 309‒314.
  • Aktaş, C. (2009). Lojistik regresyon analizi: öğrencilerin sigara içme alışkanlığı üzerine bir uygulama. Erciyes Üniversitesi Sosyal Bilimler Enstitüsü Dergisi, 1(26), 107‒122.
  • Aktaş, C. ve Erkuş, O. (2009). Lojistik regresyon analizi ile Eskişehir’in sis kestiriminin incelenmesi. İstanbul Ticaret Üniversitesi Fen Bilimleri Dergisi, 8(16), 47‒59.
  • Akyüz, G. ve Pala, N. M. (2010). Pisa 2003 sonuçlarına göre öğrenci ve sınıf özelliklerinin matematik okuryazarlığına ve problem çözme becerilerine etkisi. İlköğretim Online, 9(2), 668‒678. http://ilkogretim-online.org.tr adresinden edinilmiştir.
  • Alacacı, C. & Erbas, A. K. (2010). Unpacking the inequality among Turkish schools: Findings from PISA 2006. International Journal of Educational Developme
Daha Fazla Göster

Details

DOI 10.26650/ekoist.2018.14.28.0010

Submission : 30 Mar 2017

Early Viewed : 20 Eyl 2018

Full Text (PDF)