TY - JOUR
T1 - Estimation of children growth curve based on kernel smoothing in multi-response nonparametric regression
AU - Chamidah, Nur
AU - Saifudin, Toha
PY - 2013
Y1 - 2013
N2 - Physical children growth is measured by using anthropometric measures i.e. weight, height and head circumference. The children around two years old grow rapidly, and than decrease slowly along with increasing of children age. It means that locally model approach is more appropriate to the data. Kernel smoothing is one of estimation methods in nonparametric regression. In this paper, we study about Kernel smoothing in multi-response nonparametric regression model and apply it for estimating children up to five years old growth. The model consists of three response variables i.e. weight, height and head circumference, and age as a predictor variable. For determining optimal bandwidth for each response variable, we use cross-validation method. Based on children data in Surabaya 2010, and the 50th percentiles estimation of weight, height and head circumference versus age, we obtain the mean squared error value is 0.05583 and coefficient of determination is 99.99%. The estimation model of children growth curve based on multi-respon kernel smoothing shows fluctuation of the curve and gives mean squared error value tends to zero and coefficient of determination tends to one. These facts mean that the estimation has satisfied goodness of fit criterion.
AB - Physical children growth is measured by using anthropometric measures i.e. weight, height and head circumference. The children around two years old grow rapidly, and than decrease slowly along with increasing of children age. It means that locally model approach is more appropriate to the data. Kernel smoothing is one of estimation methods in nonparametric regression. In this paper, we study about Kernel smoothing in multi-response nonparametric regression model and apply it for estimating children up to five years old growth. The model consists of three response variables i.e. weight, height and head circumference, and age as a predictor variable. For determining optimal bandwidth for each response variable, we use cross-validation method. Based on children data in Surabaya 2010, and the 50th percentiles estimation of weight, height and head circumference versus age, we obtain the mean squared error value is 0.05583 and coefficient of determination is 99.99%. The estimation model of children growth curve based on multi-respon kernel smoothing shows fluctuation of the curve and gives mean squared error value tends to zero and coefficient of determination tends to one. These facts mean that the estimation has satisfied goodness of fit criterion.
KW - Children growth curve
KW - Kernel smoothing
KW - Multi-response
KW - Nonparametric regression
UR - http://www.scopus.com/inward/record.url?scp=84875763429&partnerID=8YFLogxK
U2 - 10.12988/ams.2013.13168
DO - 10.12988/ams.2013.13168
M3 - Article
AN - SCOPUS:84875763429
SN - 1312-885X
VL - 7
SP - 1839
EP - 1847
JO - Applied Mathematical Sciences
JF - Applied Mathematical Sciences
IS - 37-40
ER -