Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:

---

Background: In this paper, we study the Regression clustering using Dirichlet processes to achieve a comprehensive insight about the status of children and adolescence’s health based on Iranian national studies. This clustering approach, as well as removing the effects of troublous factors in clustering and obtaining more accurate analysis, estimates the number of clusters and their patterns.In the present study, subjects of the sample are clustered according to the cholesterol index and fast blood sugar (FBS) which are known as risk factors of cardiovascular disease and diabetes. Furthermore, the effect of the anthropometric index, category of age and sex on clustering are evaluated.Methods: To estimate unknown parameters of model, in a Bayesian approach, we adopt Markov chain Monte Carlo simulation methods using OpenBUGs software. In addition, to get a correct view of the number of patterns on the risk of cardiovascular disease and diabetes based on FBS and the cholesterol index, the Dirichlet Process is applied.Findings: The anthropometric index, category of age and sex as explanatory variables were entered in the model and effects of them were removed. Then, three clusters were obtained. Percentages of allocated subjects to clusters one to three are 47 %( 1310 persons), 4% (112 persons) and 49% (1366 persons), respectively. Analyzing the first and the third quantiles of the FBS and the cholesterol index in each cluster, it is revealed that the allocated subjects to the third cluster, who consist a significant part of the sample, are assumed to be at the risk of both cardiovascular disease and diabetes. Moreover, people in the first cluster have normal levels of FBS and the cholesterol index and the subjects allocated to the second cluster have dangerous level of FBS and normal level of the cholesterol index.Conclusion: In clustering of children and adolescence based on FBS and the cholesterol index, category of age, sex and the anthropometric index were known as influential factors on estimating the number of clusters and creating their structures. Also, three clusters are obtained based on this study: subjects with the safe status, subjects at the risk of diabetes and subjects at the risk of both cardiovascular disease and diabetes. Key Words: Regression Clustering, Dirichlet Process, Bayesian Approach, Food Blood Sugar, The Cholesterol Index.

Full-Text [PDF 666 kb]   (422 Downloads)    

Add your comments about this article : Your username or Email:

---