The Effect of Outliers on Linear Relationship between Mothers BMI and Infant Weight
Nwanyanwu Promise Chima & Ette Harrison Etuk
Department of Mathematics,
Rivers State University, Port Harcourt, Nigeria
Email: etuk.ette@ust.edu.ng
Corresponding Author: Ette Harrison Etuk
ABSTRACT
The purpose of this research is to study the effect of outliers on the linear regression model describing the relationship between mother’s body mass index BMI and infant weight. The weight and height of 240 mothers were collected along with the weight of their infant. A linear model was developed using simple linear regression. The model showed that the mother’s BMI has a significant effect on the weight of infant as the mother’s BMI accounted for 76.8% of the variability in infant’s weight. The regression model was tested for outliers using two methods, Data Trimming and Winsorization. Two outlier free models were developed based on the two methods used in treating the outliers and the result showed higher coefficient of determination 94.3% and 79.3% respectively and therefore more fitness than the initial model. Thus the result showed that outliers have an effect on the model as it reduced the coefficient of determination and thus its goodness of fit.
