There is an increasing concern over adult obesity which can lead to many health disorders. There is a well-defined relationship between height and weight for males and females separately, called the body mass index (BMI). BMI is the ratio of the individual's weight to the square of height. One major shortcoming of BMI is that it does not indicate health status of the person since it does not distinguish between fat and muscle mass or the distribution of fat. However, it is easy to collect and classify from large-scale nationality representative samples. Since serious health-related issues are rising due to obesity, it is important to model BMI and obesity rates. An obesity predisposing genotypes is found in 10 of individuals (Ref. 1). The observed BMI will be a combination of the underlying the BMI-type range, but conditioned by lifestyle choices. These BMI classes will react differently to different lifestyle characteristics and it is necessary to study empirically how different types react to a similar set of characteristics. Latent class modelling is commonly used in health-related studies. The latent class approach splits the popularion probabilistically into a finite number of homogeneous classes and within each class, statistical models with differing parameters are applied. Latent class modelling is based on an observed attribute called ex post and an expected value within each class. Uncovering the different features of each class is the main objective of the modelling process. The article proposes a simple way of parameterizing both class probabilities and their statistical behavior within each class by preserving ranking according to class-specific expected values.
展开▼
