Methodology

Text Box: Preparation of Data

Before any analysis can be done, the data had to be cleansed and prepared first.  This meant that data from the different sources had to be amalgamated and synthesised.  Click on each category for the method in cleansing and obtaining them.




























Boxes in blue are all attributed to each state and regional polygon.


Spatial Overlay

As each state and region in that state has attributed health, environmental, and social-economical data, each can be spatially overlaid against obesity using a range of symbology or colour schemes.  This allows the data to be visually displayed together so as to determine any spatial or statistical correlation.  Figure 1 shows Individual Median Income against obesity rates in NSW.

















Figure 1: Income overlaid against obesity rates for NSW


Statistical Analysis
Weighted overlays
After identifying the factors which contribute to obesity, the aim now is to find to what degree each will have on obesity.  Each factor will be normalised to fit a new value ranging from 0 to 100.  A new value, weighted sum, will be  calculated from weights appropriate to each factor and this will then be compared to current obesity data.  For example, the new weighted sum could consist of 30% Non School Education and 50% Income and 20% Green spaces.  Different combinations of weights will yield different weighted sums and a regression model will be calculated for each to identify the best combination.  
Regression Models
After identifying the factors which contribute to obesity from spatial analysis, Regression models were calculated for the factors against obesity.  The regression model predicts a trend for the data and will provide a ‘R squared’ value. This value represents how well the regression model is at representing the data.  The maximum value for this is 1 which signifies that the regression model directly fits the data.  Regression models were also calculated for weighted overlays to identify the overlay which best reflects the obesity data, i.e. the highset R Squared value.

Obesity in Australia :

A study of contributing socio—economical factors in using Spatial Analysis 

Author: Lucy Chen

Supervisor  : Samsung Lim