New Statistical Tools For Modeling Children's Body Mass Index
Juntunen, Janni (2020)
Juntunen, Janni
2020
Laskennallisen suurten tietoaineistojen analysoinnin maisterikoulutus, FM (engl) - Master's Degree Programme in Computational Big Data Analytics
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2020-09-10
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202007156359
https://urn.fi/URN:NBN:fi:tuni-202007156359
Tiivistelmä
Overweightness and obesity among children have been increased considerably in the last decades. Thus, children’s Body Mass Index (BMI) curves have been examined in order to identify the children with a risk of obesity. For instance, group-based modeling for BMI have been used to group children based on their BMI measurements. An important concept has also been adiposity rebound (AR), which is the second rise of the BMI curve, occurring approximately at the age of six. Previous research has demostrated that there seems to be an association between an earlier AR and later obesity.
The aim of this thesis is to introduce a new statistical tool, a semiparametric mixture regression model, for group-based modeling for children’s BMI, and use the tool to identify groups of children who are potentially obese in later life. Furthermore, this thesis aims to estimate the age at AR in these groups with different estimation methods and compare these methods. The estimation methods used were cubic smoothing splines with different smoothing options and a third order polynomial regression model. All the analyses were conducted with statistical software R, and the analyses were done separately for boys and girls. The data included height and weight measurements from 0 to 15 years from 2854 Finnish children, born in Pirkanmaa in 2003.
With the semiparametric mixture regression model, the children could be divided into four groups, called trajectories, on the basis of their BMI measurements in different ages. Among boys and girls the uppermost trajectories could lead to overweight in adolescence. These trajectories also had higher BMI in the last measurements (ages 12 to 15). The results from the different estimation methods showed that AR occurred earlier in the the uppermost trajectories and later in the undermost trajectories. Although the estimated ages at AR differed between the methods, the fnal results described above were the same.
These results suggest that potentially obese children could be identifed with the semiparametric mixture regression model. Additionally, the results propose that AR occurs earlier among the children with a risk of obesity. Concerning the different methods for estimating the age at AR, the consistent fnal results indicate that similar conclusions can be achieved with different methods, at least if the target is in comparing children or groups of children.
The aim of this thesis is to introduce a new statistical tool, a semiparametric mixture regression model, for group-based modeling for children’s BMI, and use the tool to identify groups of children who are potentially obese in later life. Furthermore, this thesis aims to estimate the age at AR in these groups with different estimation methods and compare these methods. The estimation methods used were cubic smoothing splines with different smoothing options and a third order polynomial regression model. All the analyses were conducted with statistical software R, and the analyses were done separately for boys and girls. The data included height and weight measurements from 0 to 15 years from 2854 Finnish children, born in Pirkanmaa in 2003.
With the semiparametric mixture regression model, the children could be divided into four groups, called trajectories, on the basis of their BMI measurements in different ages. Among boys and girls the uppermost trajectories could lead to overweight in adolescence. These trajectories also had higher BMI in the last measurements (ages 12 to 15). The results from the different estimation methods showed that AR occurred earlier in the the uppermost trajectories and later in the undermost trajectories. Although the estimated ages at AR differed between the methods, the fnal results described above were the same.
These results suggest that potentially obese children could be identifed with the semiparametric mixture regression model. Additionally, the results propose that AR occurs earlier among the children with a risk of obesity. Concerning the different methods for estimating the age at AR, the consistent fnal results indicate that similar conclusions can be achieved with different methods, at least if the target is in comparing children or groups of children.