المؤلفون
.
الملخص
This paper involves applied a binary logistic regression as a traditional method for modeling and analyzing the effect of a set of explanatory variables on anemia (including two types, chronic and acute anemia), which represents a categorical binary response variable in some of Iraqi hospitals. The analysis was considered that females are the reference level. The Bayes method was applied as a modern method to estimating the parameters of a binary response regression model. Both the non-information and information prior probability distributions were used to find the Bayesian estimates. The results of the Bayesian estimates were compared according to probability function. A comparison was also made between the traditional method and the results obtained from the Bayes method. The AIC, SC, and 2 Log L as a model selection criterion were used. Both the Receiver Operating Characteristic and the classification table were used also to identify the model's accuracy and classification capability. The paper found that the parameters estimates of the binary logistic regression has similar to the non-informational prior distribution with a very slight superiority of the latter method according to the standard deviations of the estimated parameters as well as the model selection criterion. The analysis showed a clear superiority of the Bayes method according to the information prior distribution based on estimates of the logistic regression parameters on the other estimates. The analysis showed the high classification capacity of the Bayes method according to the prior information distribution. One of the most important variables affecting anemia in males (considering that females are the reference level), is sex and the Reticulocyte count as well as iron deficiency in the blood.
الكلمات الرئيسة
)