Akademik Veri Yönetim Sistemi
Turkiye Klinikleri Journal of Biostatistics, cilt.9, sa.1, ss.23-34, 2017 (Hakemli Dergi)
Objective: Entropy Optimization Methods (EOM) have important applications,
especially in statistics, economy, engineering, survival data analysis and etc. There are several
examples in the literature that known statistical data do not conform to theoretical distributions, however do conform the entropy optimization distributions well. In the present study,
survival data of male patients with localized cancer of a rectum diagnosed in Connecticut from
1935 to 1944 is analyzed by using Generalized Entropy Optimization Methods (GEOM) in the
form of the MinMinxEnt and the MaxMinxEnt methods. Material and Methods: The
MinMinxEnt and the MaxMinxEnt methods have suggested distributions in the form of the
MinMinxEnt, the MaxMinxEnt distributions which are closest to statistical data and furthest
from statistical data in the sense of Kullback-Leibler measure, respectively. Results: The results
are acquired by using statistical software MATLAB. The performances of MinMixEnt and
MaxMinxEnt methods are established by Chi-Square, Root Mean Square Error (RMSE) and
Kullback-Leibler criteria. It is shown that
is better than
distribution in the sense of Kullback-Leibler measure to mentioned data. Furthermore, in the
sense of RMSE criteria
distribution is more suitable for statistical data than
distribution. These results are also corroborated by graphical representation.
Conclusion: In this study, it is shown that
and
distributions
more successfully represent Survival Data. Our investigation indicates that GEOM in survival
data analysis yields reasonable results.