Journal Article

Hierarchical Modeling of The Risk Factors of Hypertension Cases in Some Selected Hospitals in Kebbi State Through The Use of Log-Linear Analysis

Abstract

This study aimed to understand the interactions and contributions of various risk factors of hypertensionin some selected hospitals in Kebbi State by fitting log-linear models, testing for main and interactioneffects, evaluating goodness-of-fit measures, and making inferences about the relationships among thevariables. The data set consisted of 405 valid cases of four risk factors. Cell counts and residualsexamined the match between observed and expected frequencies for different combinations of riskfactors. SPSS software was used for the analysis. There exists a strong fit of the log-linear models to thedata. Extremely small p-values indicate a high level of statistical significance and a perfect fit. The KWayand Higher-Order Effects analysis explores the significance of different orders of effects. The resultsdemonstrate that first, second, and third-order effects have high statistical significance, while fourth-ordereffects are considered insignificant. Backward elimination statistics are employed to refine the model byremoving non-significant variables. The iterative procedure helps identify the most parsimonious andinterpretable model. The goodness-of-fit tests conducted on the final model confirm its suitability, withboth the likelihood ratio test and the Pearson test yielding high p-values. These results indicate a strong fitbetween the model and the observed data.

Download pdf