Amaç: Sağlık alanında tüm hastalardan aynı zaman noktalarında ölçüm yapmak her zaman mümkün değildir. İşlemler her bireyde eşit aralıklarla yapılsa bile farklı zamanlarda başlayabilir. Bazen de işlemin bir gereksinimi olarak değişik zaman aralıkları ile ölçümler yapılır. Bu tür tekrarlanan ölçümlü verilerin analizinden doğru sonuçlar elde edebilmek için doğru bir kovaryans modeline ihtiyaç vardır. Tekrarlanan ölçümlü veri analizinde, veriye en uygun varyans kovaryans yapısının belirlenmesi doğru çözüm için esastır. Bu çalışmanın amacı tekrarlanan ölçümlü verilerin analizinde varyans kovaryans modeli hatalı seçildiğinde oluşabilecek farklılıkları saptayarak nihai kararı ne kadar etkilebileceğini test etmektir. Gereç ve Yöntemler: Çalışmada varyans kovaryans yapısına uygun veriler SAS paket programı kodlama sistemi içerisinde simülasyon ile türetilerek gerçek varyans kovaryans yapısı bilinen bu veri setleri farklı varyans kovaryans yapıları ile analiz edilerek, hatalı varyans kovaryans yapısı seçiminin analiz üzerine etkisi incelendi. Bulgular: Çalışmada tekrarlanan ölçümlü verilerin analizi için en uygun varyans kovaryans yapısı, BIC değeri minimum olan olan Yapılanmamış (UN) varyans kovaryans modeli olarak belirlendi. Veri Yapılanmamış (UN) varyans kovaryans modeli ile analiz edildiğinde ilaç etkisi önemli olarak bulundu. GLM ile analiz edildiğinde ilaç etkisi önemsiz olarak bulundu. Sonuç: Eksik verili tekrarlanan ölçümlerin analizinde GLM çözüm tekniğinin kullanılması kararların hatalı olmasına sebebiyet verebilir. Bu sebeple eksik verili tekrarlanan ölçümlerde MIXED çözüm tekniği kullanılması önerilmektedir. Aksi halde gerçekçi olmayan kararların verilmesi söz konusudur.
Anahtar Kelimeler: GLM; MIXED; Tekrarlanan Ölçüm; Kovaryans Modelleri
Objective: It is not always possible to make measurements at the same time points from all patients in the field of health. Transactions may begin at different times, even if they are performed at equal intervals in each individual. Occasionally, measurements are made at different time intervals as a requirement of the process. In order to obtain accurate results from the analysis of such repeated measurement data, an accurate model of covariance is needed. In repeated data analysis, determining the most appropriate variance covariance structure for data is essential for the correct solution. The aim of this study is to test the effects of repeated measurement data and determine the differences that may occur when the variance covariance model is selected incorrectly and to test the final decision. Material and Methods: In this study, the data on the variance covariance structure and SAS package program coding system were analyzed by using different variance covariance structures and the effect on the analysis of the selection of incorrect variance covariance structure was investigated. Findings: In the study, the most appropriate variance covariance structure for the analysis of repeated measured data was determined as Unstructured (UN) variance covariance model with minimum BIC value. When the data were analyzed by Univariate (UN) varicence covariance model, the drug effect was found to be significant. When analyzed by GLM, the drug effect was insignificant. Conclusion: The use of the GLM solution technique in the analysis of incomplete data may result in erroneous decisions. For this reason, it is recommended to use MIXED solution technique for repeated measurements. Otherwise, unrealistic decisions are made.
Keywords: GLM; MIXED; Repeated Measures; Covariance Structures
- Wolfinger RD. Covariance structure selection in general mixed models. Communications in Statistics-Simulation and Computation. 1993;22(4):1079-106. [Crossref]
- Littell RC, Milliken GA, Stroup WW, Wolfinger RD, Schabenberger O. SAS for Mixed Models. 2nd ed. Cary, NC, USA: SAS Institute Inc.; 2006. p.795.
- Kincaid C. Guidelines for Selecting the Covariance Structure in Mixed Model Analysis. SAS Conference 30 Proceedings: SAS Users Group International. Philadelphia, Pennsylvania: SAS Institute Inc.; 2005. p.8.
- Wolfinger RD. Heterogeneous variance: covariance structures for repeated measures. Journal of Agricultural, Biological, and Environmental Statistics (JABES). 1996;1(2):205-30. [Crossref]
- Weiss RE. Modeling Longitudinal Data. 1st ed. New York: Springer-Verlag; 2005. p.432.
- Anderson JE. (2004). PROC MIXED and Repeated Measures. Erişim:[http://cda.morris.umn.edu/~anderson/math4601/gopher/SAS/longdata/structures.pdf]. Erişim Tarihi: 09.10.2018.
- Wolfinger RD. An example of using mixed models and PROC MIXED for longitudinal data. J Biopharm Stat. 1997;7(4):481-500. [Crossref] [PubMed]
- Fernandez G. Model Selection in PROC MIXED- A User-friendly SAS Macro Application. SAS Conference Proceedings: SAS Global Forum 2007, Orlando, Florida. 2007. p.20. [Crossref]
- Anderson DR, Burnham KP, White GC. Comparison of akaike ınformation criterion and consistent akaike ınformation criterion for model selection and statistical ınference from capture- recapture studies. Journal of Applied Statistics. 1998;25(2):263-82. [Crossref]
- Moser EB. Repeated Measures Modeling With PROC MIXED. SAS Conference Proceedings: SAS Users Group International 29. Montreal, Canada. 2004. p.19.
- Brown H, Prescott R. Applied Mixed Models in Medicine. 2nd ed. New York: John Wiley and Sons; 2006. p.476. [Crossref]
- Fitzmaurice GM, Laird NM, Ware JH. Applied Longitudinal Analysis. 1st ed. Hoboken, NJ: John Wiley & Sons Inc.; 2004. p.506.
- Wolfinger RD, Chang M. Comparing the SAS® GLM and MIXED Procedures for Repeated Measures. SAS Conference Proceedings: SAS Users Group International 20. Orlando, Florida. 1995. p.11.
- Hamer RM, Simpson PM. Mixed-Up Mixed Models: Things That Look Like They Should Work But Don't, and Things That Look Like They Shouldn't Work But Do. SAS Conference 25 Proceedings: SAS Users Group International. Indianapolis, Indiana. 2000. p.8.
- Yarandi HN. Comparison of PROC MIXED and PROC GLM for Analysis of Repeated Measures Data. Proceding of the 28 Annual SAS Users Group Conference. Florida. 2003. p.6.
- Bushnell W, Steiner M. Use of PROC MIXED in the Analysis of Repeated Measures Data from a Clinical Trial in Obsessive Compulsive Disorder. SAS Conference 22 Proceedings: SAS Users Group International. San Diego, California. 1997. p.6.
- Johnson M. The Effect of Missing Data on Repeated Measures Models. SAS Conference 24 Proceedings: SAS Users Group International. Miami Beach. 1999. p.262-7.
- Howell DC. Mixed Models For Repeated (Longitudinal) Data. 2008. Erişim:[http://www.uvm.edu/~dhowell/StatPages/]. Erişim Tarihi: 01.04.2019
- Wolfinger RD, Chang M. Comparing the SAS® GLM and MIXED Procedures for Repeated Measures. SAS Conference 24 Proceedings: SAS Users Group International. Orlando, Florida. 1995. p.11.
- Johnson DE. An Introduction to the Analysis of Mixed Models. SAS Conference Proceedings: SAS Users Group International 28. Seattle, Washington. 2003. p.4.
- AL-Marshadi AH. The new approach to guide the selection of the covariance structure in mixed model. Research Journal of Medicine and Medical Sciences (RJMMS). 2007;2(2):88-97.
- Gómez S, Torres Cárdenas V, Mesa MY, Sardi-as Y, Rodríguez Y, Herrera M, Calvo Rodríguez YR. Application of the linear mixed and generalized mixed model as alternatives for analysis in experiments with repeated measures. Cuban Journal of Agricultural Science. 2019;53(1):7-12.
- Huang K, Walker CA. Comparisons of statistical models for growth curves from 90-day rat feeding studies. Archives of Toxicology. 2019;93(8):2397-408. [Crossref] [PubMed]
- McFarquhar M. Modelling group-level repeated measurements of neuroimaging data using the univariate general linear model. Front Neurosci. 2019;13(352):1-13. [Crossref] [PubMed] [PMC]
.: Process List