Categorising variables in medical contexts

[摘要] Many medical studies involve modelling the relationship between an outcomevariable and a series of one or more continuous/interval scaled discrete explanatoryvariables. It is common practice in many of these studies for some, or indeed all, ofthe continuous/interval scaled discrete explanatory factors to be incorporated into theanalysisi n a categorisedo r groupedf orm.One of the main reasons for adopting this methodology is that it will simplifythe interpretation of results for clinicians and hopefully patients. It is often easier tointerpret conclusions based on an explanatory variable with two or three levels (i. e.categorisations) than from a continuous/interval scaled discrete explanatory. Themain drawback with this technique is in identifying the categorisation points. Oftenpreconceived and/or historical grounds are the determining factor used to decide thelocation of these categorisation points. However, this may not give rise to sensible orjustifiable locations for such points for a given application.This thesis will consider the analysis of data from various types of medicalstudy and, by applying non-parametric statistical methodology, provide alternative,more logical rationale for identifying categorisation points. The analysis willconcentrate on data from three specific types of medical study -a cohort study with abinary outcome, a matched case/control study and survival analysis.In a cohort study with a binary response the standard methodology of logisticregression will be applied and extended using a non-parametric logistic approach toidentify potential categorisation points. As a further extension consideration will begiven to the more formal methodology of examining the first derivative of theiiresultant non-parametric logistic regression to provide the location of categorisationpoints.In matched caselcontrol studies the standard technique used for analysis isconditional logistic regression. The theory and application of this model will bediscussed before considering two new, alternative, non-parametric approaches toanalysing matched case/control studies with an interval scaled discrete explanatoryvariable. The proposedn on-parametrica pproachesw ill be testedt o investigatet heirusefulness in identification of categorisations for the explanatory variable. Possibleextensionst o thesea pproachesto incorporatea single continuouse xplanatoryv ariablewill be discussed. In order to compare the two non-parametric approaches asimulation study will be carried out to investigate the power of these approaches.Finally, consideration will be given to the analysis of survival data. Initially,the standard methodologies of the Kaplan and Meier estimator in the absence ofexplanatory variables and Cox's Proportional Hazards model to incorporateexplanatory variables will be discussed. A more detailed examination of threealternative methods for analysing survival data in the presence of a single continuousexplanatory will be carried out. Each of the methods will be applied in turn to asurvival analysis problem to investigate if any categorisationsc an be identified for asingle continuous explanatory variable. Further simulations will be undertaken tocompare the three methods across a variety of scenarios.

[发布机构] University:University of Glasgow;Department:School of Mathematics and Statistics

[关键词] QA Mathematics