Comparison of Log-Linear Analysis and Correspondence Analysis in Two-Way Contingency Tables: A Medical Application

Abstract

Objective: The aim of the study was to examine log-linear analysis and correspondence analysis by an application with regard to advantages and disadvantages of the two methods. Material and Method: In this study, we used an artificial data set, which is expanded without changing its nature, from a study conducted in Medical Faculty of Baskent University, Infectious Disease and Clinical Microbiology Department. Relations and interactions between variables and among subcategories have been investigated with log linear and correspondence analysis. Results: It has been shown that when the assumptions of log-linear analysis were not available, better understanding of the data set could be obtained by correspondence analysis (CA). However, conclusions about the data may not be generalized in a confidence interval to the population by CA. In addition, in log-linear analysis, it is possible to make inferences about the population on the basis of sample data. The other important result obtained from the study was that some of the relations between the same variable categories could be detected by CA, which is not possible with log-linear analysis. Conclusion: As a result, we suggest the complementary use of log-linear analysis and correspondence analysis for detailed results.

Keywords

• Correspondence analysis, log-linear analysis, inertia, saturated models, unsaturated model

Comparison of Log-Linear Analysis and Correspondence Analysis in Two-Way Contingency Tables: A Medical Application

Öz

Kaynakça

1. Agresti A. An Introduction to Categorical Data Analysis. 1st ed. USA: John Wiley & Sons; 1996;145-50.
2. Knoke D, Burke PJ. Log-Linear Models. 1st ed. USA: Sage Publi- cations; 1980;8-17.
3. Greenacre M, Blasius J. Correspondence Analysis in the Social Sciences. 1st ed. London; Academic Press; 1994;3-111.
4. Hoffman DC, Franke GR. Correspondence Analysis: Graphical Representation of Categorical Data in Marketing Research. Jour- nal of Marketing Research 1986;23:213-27.
5. Van der Heijden PGM, de Falguerolles A, de Leeuw J. A Com- bined Approach to Contingency Table Analysis Using Correspon- dence Analysis and Loglinear Analysis. Applied Statistics 1989; 38: 249-92.
6. MINITAB, Version 14.0, MINITAB Inc. 2006.
7. NCSS: Number Cruncher Statistical System, Version 2004. J Hin- tze, Kaysville, UT, 2003, in Basic and Clinical Biostatistics, Daw- son B, Trapp RG, New York; Lange Medical Books/McGraw-Hill; 2004. pp. ix, x.
8. SPSS: Statistical Program for Social Sciences, Version 11.5, Spss Inc. Chicago, II, 2003.

9. Theus M, Lauer SRW. Visualizing Loglinear Models. Journal of Computational and Graphical Statistics 1999;8:396-412.
10. Andersen EB. Introduction to the Statistical Analysis of Categori- cal Data. 1st edition. Germany: Springer Verlag 1997;217-24.
11. Yazici, A.C. The application and the Interpretation of the Cor- respondence Analysis in Discrete Data Obtained from Biological Events. PhD Thesis. Ankara University, Ankara 2003.
12. Hawkins DM. Topics in Applied Multivariate Analysis. 1st edition. USA: Cambridge University Press; 1982;183-206.
13. EscoŞ er B, Pages J. Presentation of Correspondence Analysis and Multiple Correspondence Analysis with the Help of Examples, In: Devillers J and Karcher W, eds. Applied Multivariate Analysis in SAR and Environmental Studies, Netherlands: Kluwer Academic Pub;1991;1-31.
14. Friendly M. Mosaic Displays for Multi-Way Contingency Tables. Jasa 1994;89:190-200.
15. Greenacre M, Hastie T. The Geometric Interpretation of Corre- spondence Analysis. Journal of the American Statistical Associa- tion 1987;82:437-47.
16. Michailidis G, De Leeuw J. The GiŞ System of Descriptive Multi- variate Analysis. Statistical Science 1998;13:307-36.
17. GiŞ A. Nonlinear Multivariate Analysis, New York: John Wiley &Sons; 1990;265-72.
18. Greenacre M. Correspondence Analysis in Practice. 2nd edition. London: Academic Press; 2007;97-144.
19. Panogiotakos DB, Pitsavos C. Interpretation of Epidemiological Data Using Multiple Correspondence Analysis and Log-linear Models. Journal of Data Science 2004;2:75-86.
20. Van der Heijden PGM, Worsley KJ. Comment on “Correspon- dence Analysis Used Complementary to Loglinear Analysis”. Psy- chometrica 1988;53:287-91.
21. Moser EB. Exploring Contingency Tables with Correspondence Analysis. Bioinformatics 1989;5:183-9.