| Summary: | 碩士 === 銘傳大學 === 醫療資訊與管理學系健康產業管理碩士班 === 107 === Obesity is one of the global public health problems and it is associated with an increased risk of many chronic diseases. Furthermore, the quality of life may also be affected by obesity. In the past, most of the predictive factors in the studies performed the regression analysis targeted at the mean value of Body Mass Index (BMI). However, such methods are limited because the impact of each predictive factor over BMI may change with different quantiles. If such a phenomenon occurs, the use of quantile regression analysis will be more appropriate than the traditional regression analysis.
In this study, the longitudinal method of repeated observation is conducted by using the health data obtained from a health examination center of adults aged over 20 years from 2010 to 2015. After the missing values of diabetes mellitus and variables are excluded, there is a total of 140,904 (63,817 persons) pieces of data included in the analysis. Mixed quantile regression analysis is used to explore the relationship between age, marital status, educational level, family income, smoking, drinking, betel nut chewing, sleeping time, exercise status, vegetarianism, and self-perceived health condition at different BMI quantiles. After the assumption of correlation distribution is given, the mixed quantile regression parameter estimation is carried out, and lqmm packages of R-studio are used for analysis.
The study finds that the age, education level, and family income have different effects on BMI of male and female. The effects of male age on BMI are particularly obvious on the low quantiles and high quantiles. At the 10th quantile, the higher the age, the higher is the BMI. However, at the 90th quantile, the opposite results emerge: the younger the age, the higher is the BMI. Among the whole BMI distribution, BMI for females increase with their age. The educational background is not significantly correlated with BMI for male, but BMI for female drop significantly when the educational background is higher. The higher the family income for males, the higher is the BMI. The family income for males exert more effects on BMI at low quantile and high quantile, showing a U-shaped trend. Among the entire distribution of BMI, that for males and females decreases when self-perceived health condition is better. Besides, the male smoking regularly and female having spouse and being vegetarian exert different effects on BMI at its different quantiles, indicating the different effects on BMI among under-weight, normal weight, and over-weight individuals. This study finds that for different genders the effect by each variable upon BMI varies at different quantiles. Thus, the differences existing in the different distribution of BMI population should be considered when the influence factors of obesity are deduced.
|
|---|
