Exploratory analyses of caries data

Some recent developments in statistical technique are applied to exploring published data of caries experience in 5-year old children. The methods are: robust regression and identification of outliers; analysis of compositional data derived from the dmf index; plotting compositional data using a ternary (or Ferrers) diagram, with a 96 per cent contour and principal axes; data-smoothing applied to 3-dimensional graphs of prevalence, severity and treatment response to caries. A strong relationship was found between prevalence and severity of caries, and 16 outlying observations were identified. The smoothed data demonstrated an insufficient treatment response to caries. The role of such analyses is discussed and recommendations made.