A Note on the Application of Analysis of Variance in Medical Research

Abstract

Some of the data related to bioscience are socioeconomic characteristics which are associated with different types of communicable or non-communicable diseases. For example, age, height, weight, gender, marital status, religion, education, occupation, income, food habit, working conditions, smoking habit, and utilization of time are associated with many noncommunicable diseases. This phenomenon was observed in many studies, home and abroad [ 1-20], especially these characteristics were found associated with level of obesity, prevalence of diabetes and prevalence of other non-communicable diseases. The level of obesity is measured by the ratio of body weight (in kg) to square of height (in m2). This measurement is known as body mass index (BMI) and it is a quantitative measurement. The study of association of this variable with any of the social characteristics is not possible if the variables are not classified into groups. For example, level of income and level of obesity are significantly associated [21]. Here level of obesity depends on level of income and this dependency can be studied if the values of both the variables are measured by numerical figures. In this type of analysis, the main objective is to verify whether the values of BMI vary with the variation of income and for which level(s) of income the BMI is significantly different. This type of information is not available from the study of association. This is possible if analysis of variance, one-way or multi-way, is applied. The analysis of variance helps to distinguish between the average values of BMI for any two or more levels of income. In this note one-way analysis of variance was presented to distinguish the level of BMI of children and adults of Bangladesh with respect to some social characters.