Re-Weighted Least Squares: The Best Negative Binomial Regression Methods In Determining The Congenital Anomalies�? Risk Factors

Abstract

Ahmed Hamza Abood, Suhair Khatan Ismail, Adnan Fadhil Toma Sabah

Nowadays, the prevalence of congenital anomalies (CAs) is increasing. Our study aimed to assess the best method among various negative binomial regression methods that identifies the possible risk factors of CAs among infants attending the Pediatric Hospital in Karbala government, Iraq using mean square error MSE and Determination Coefficient R2 as a Comparative criterions. We did a cross-sectional retrospective study in which a review of the record checklists of a 257 neonates admitted in the hospital over a three year period (January 2016–December 2019). Entering and analyzing of the data were performed using statistical program SPSS version 21. The results showed that firstly, the best distribution for CAs risk factors is negative binomial. Secondly, IRLS has higher R2 and lower MSE values than PLS. Thirdly, the variables (Mother health, Degree of parents kinship, type and present births) were significant, with a value of less than 0.05 in all methods. We concluded that IRLS is the best negative binomial regression method that determines the congenital anomaliesÒ† risk factors.