Multiple Linear Regression Model based on Principal Component Scores to Study the Relationship between Anthropometric Variables and BP Reactivity to Unsupported Back in Normotensive Post-graduate Females

Arterial Blood pressure (BP) is influenced by numerous factors. Guidelines on measurement of BP contain recommendations on the position of the back by advising that the subject should sit with back supported. The objectives of this work are: i) to determine whether unsupported back during BP measurement in sitting position has significant effect on BP, ii) to determine the relationship between some anthropometric variables (age, height, weight, BMI and arm circumference) and BP reactivity to unsupported back using Principal component analysis (PCA) based multiple linear regression (MLR) analysis. According to the results of paired t-test, unsupported back does not have significant effect on systolic BP (SBP) measurements, P =0.09, but statistically significant effect on diastolic BP (DBP) measurements P<0.001. PCA based MLR analysis was used then to determine the relationship between DBP reactivity and anthropometric variables, as multicollinearity exist between the anthropometric variables. Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy (0.63) and Barlett’s test of sphericity (chi square = 231.01, DF=10, P<0.0001) proved that the data were fit for application of PCA. Results of PCA demonstrated that out of 5 principal components (PCs) only first 4 (PC1, PC2, PC3, PC4), each explaining more than 5% of variance, were used for MLR analysis. Results of PCA based MLR analysis showed that PC3 (P<0.001) and PC4 (P=0.014) had significant effect on BP reactivity, and accounted for 64.2% of total variation in the BP reactivity to unsupported back. This assessment presents the importance and advantages poses by PCA based MLR analysis to study the relationship between variable sets in research studies. Key-Words: Normotensive post-graduate females, unsupported back, BP reactivity, anthropometric variables, multicolinearity, multiple linear regression, principal component analysis.