LMS Method教程.ppt

Using theLMS Method with lmsChartMaker and lmsGrowth Sanja Stanojevic UCL Institute of Child Health Growth Charts: Centiles Centile reference charts are used in medicine to observe clinical measurements on individual patients in the context of population values If the subject’s centile value is atypical (e.g. below the 3rd, above the 97th or outside ± 2 SDS) this may indicate an underlying pathological condition Centiles can also be used to track an individual’s measurements over time LMS Reference centile curves show the distribution of a measurement (e.g. height, weight etc.) as it changes according to an independent variable (e.g. age) LMS Using penalised likelihood the three curves (LMS) can be fitted as cubic splines by non-linear regression, and the extent of smoothing required can be expressed in terms of smoothing parameters of equivalent degrees of freedom (edf) Summary so far Describes how the selected centiles for a particular measurement change when plotted against an independent variable (time, age, etc.) Centiles are drawn to follow the distribution of the data, allowing for roughness (trade-off between smoothness and goodness of fit) Smooth centile curves based on the Box-Cox power transformation family adjusted for non-linearity, heteroscedasticity and skewness of data Software Download lmsChartMaker and lmsGrowth from /tjcole Use lmsChartMaker to construct centile curves from reference data Use lmsGrowth to assess measurements against an existing growth reference Using lmsChartMaker Importing Data Columns are separated by a “box” in the first row Fill in the appropriate column numbers The predictor variable (age) is the 4th column, the dependent variable (FVC) is the 6th, and the selector variable (sex) is the 3rd column Renaming Variables Selecting Groups (if you want to analyse different groups separately) Modelling Data Selecting Models Choosing the Best Model 4 units of deviance for 1 edf is approximate to 5% significance Deviance changes