Cubic Smoothing Spline on some Selected Blood Pressure Variables
Keywords:
Generalized Additive Model, Categorical Covariates, Metrical Covariates, Smoothing Spline, Blood PressureAbstract
Generalized Additive Model has become an elegant and practical option in modelling non-linear and linear effects
of covariates as well as the non-Gaussian response variable. This study considered modelling Blood Pressure (BP)
using data with two levels of BP (abnormal and normal) and eight predictors which have both linear and non-linear
effects. The non-parametric functions were estimated in a flexible manner using cubic smoothing spline in an
iterative method called the Back-fitting algorithm. The Cubic smoothing spline was applied to the metrical
covariates (Age and BMI), which gave significant results (p < 0.0001 and 0.0082 respectively) compared to the
linear fit which was not significant. The empirical findings of this study have established that BMI and Age have
significant non-linear effect while sex and cholesterol level have significant linear effect on BP.
