Meta analizinde heterojenlik ölçütlerinin karşılaştırılması
Date
2024-05-20
Authors
Toluk, Özlem
Journal Title
Journal ISSN
Volume Title
Publisher
Bursa Uludağ Üniversitesi
Abstract
Meta analiz farklı araştırmacılar tarafından, farklı zaman, örneklem ve yerlerde yapılmış olan çalışmaların verilerinin birleştirilip analiz edilerek klinik karar sürecine destek olmayı sağlayan önemli bir yöntemdir. Tez çalışmasında meta analizine alınan çalışmalar homojen ve heterojen olarak türetilmiştir. Hastalık-etken olasılıkları P=0,5; 0,6; 0,7; 0,8 ve 0,9 alınarak, 1.000.000 birimden oluşan varsayımsal anakütleler oluşturularak, 1.000 defa örnek çekilmiştir. Farklı P olasılıklarına sahip varsayımsal anakütlelerden farklı örneklem büyüklükleri (n) ve farklı çalışma sayılarına (k) göre alınan örneklemler ile meta analizi yapılmış, 𝜏2, 𝐼2, 𝑅𝑏 ve H heterojenlik ölçütlerinin performansları karşılaştırılmıştır. Tezin ikinci kısmında literatürde geçerli olan 𝐼2 ölçütüne göre çalışmalarda az, orta ve yüksek heterojenlik oluşturulmuştur. Heterojen çalışmalar farklı örneklem büyüklükleri, farklı çalışma sayıları ve farklı hastalık-etken olasılıklarına göre oluşturulmuştur. Heterojen çalışmalara meta analiz yapılarak 𝜏2, 𝐼2, 𝑅𝑏 ve H heterojenlik ölçütlerinin performansları karşılaştırılmıştır. Ölçütlerin performanslarının yanında homojen ve heterojen çalışmalar için Cochran’ın Q istatistiğine yönelik performans değerlendirmesi yapılmıştır. Yüksek örneklem büyüklüklerinde ve yüksek çalışma sayılarında en iyi performansı 𝜏2 göstermiş onu H ölçütü izlemiştir. 𝐼2 ve 𝑅𝑏 heterojenlik ölçütleri genel olarak küçük çalışma sayıları ve küçük örneklem büyüklüklerinde iyi performans göstermişlerdir.
Meta-analysis is a crucial method that supports the clinical decision process by combining and analyzing data from studies conducted by different people, at different times, samples, and places. The studies included in the meta-analysis in the dissertation study were derived as homogeneous and heterogeneous. Disease-event probabilities by taking P=0.5; 0.6; 0.7; 0.8 and 0.9, hypothetical populations consisting of 1,000,000 units were created and samples were drawn 1,000 times. A meta-analysis was conducted with samples taken from hypothetical populations with different P probabilities, with different sample sizes (n) and different number of studies (k), then performances of the heterogeneity measures 𝜏2, 𝐼2, 𝑅𝑏, H were compared. Heterogeneous studies were generated according to different sample sizes, different number of studies and different disease-event probabilities. Meta-analysis was performed on heterogeneous studies and the performances of 𝜏2, 𝐼2, 𝑅𝑏 and H heterogeneity measures were compared. In addition to the performance of the measures, performance evaluation was performed for Cochran's Q statistic for homogeneous and heterogeneous studies. The 𝜏2 performed high performance in high sample sizes and high number of studies and followed by the H measure. The 𝐼2 and 𝑅𝑏 measures generally performed well with small sample sizes and small number of studies.
Meta-analysis is a crucial method that supports the clinical decision process by combining and analyzing data from studies conducted by different people, at different times, samples, and places. The studies included in the meta-analysis in the dissertation study were derived as homogeneous and heterogeneous. Disease-event probabilities by taking P=0.5; 0.6; 0.7; 0.8 and 0.9, hypothetical populations consisting of 1,000,000 units were created and samples were drawn 1,000 times. A meta-analysis was conducted with samples taken from hypothetical populations with different P probabilities, with different sample sizes (n) and different number of studies (k), then performances of the heterogeneity measures 𝜏2, 𝐼2, 𝑅𝑏, H were compared. Heterogeneous studies were generated according to different sample sizes, different number of studies and different disease-event probabilities. Meta-analysis was performed on heterogeneous studies and the performances of 𝜏2, 𝐼2, 𝑅𝑏 and H heterogeneity measures were compared. In addition to the performance of the measures, performance evaluation was performed for Cochran's Q statistic for homogeneous and heterogeneous studies. The 𝜏2 performed high performance in high sample sizes and high number of studies and followed by the H measure. The 𝐼2 and 𝑅𝑏 measures generally performed well with small sample sizes and small number of studies.
Description
Keywords
Meta analizi, 𝐼2 heterojenlik ölçütü, 𝑅𝑏 heterojenlik ölçütü, H heterojenlik ölçütü, Tau2 heterojenlik ölçütü, Simülasyon, Meta analysis, 𝐼2 heterogeneity measure, 𝑅𝑏 heterogeneity measure, H heterogeneity measure, Tau2 heterogeneity measure, Simulation