Analyzing association and repeated measures data

Abstract

Sir, We read with great interest, the series on research methodology in dentistry[1 2] by Krithikadatta J, Valarmathi S. We congratulate the authors on the detailed description given on the how's and whys of the research technique and also the common errors that are observed in the representation of the data. We would like to add to their wonderful article a few key analysis techniques that are commonly encountered in a dental research. In cases where a patient is assessed more than twice for a variable, for example the flow of saliva is assessed at various stages of radiotherapy, Repeated Measures ANOVA or a General Linear Model of analysis is advocated. This kind of data is also obtained when a treatment has multiple stages, for example leukoplakia, oral sub mucous fibrosis etc [Figure 1].[3] Figure 1 Test for Repeated Measures In cases where we are looking for analysis of association of two parameters, correlation is commonly used, although they do not relate the cause to effect of the parameters concerned. For normally distributed data, Pearson's correlation is used, and for skewed data or ordered categorical data, Spearman's correlation is used. This gives the correlation coefficient “r,” which gives the level of correlation, and “r2” value gives the percentage of cases, which are showing the association, both of which has to be reported [Figure 2]. This gives a one-sight indication of the confounding and interaction between the parameters.[3] Figure 2 Tests for correlation The importance of the graph cannot be over looked. Normality is represented by a histogram, data can be represented by various types of bar chart, box plot, line, error chart, scatter diagram etc., which reduce the number of tables and also enhance the understanding of the outcome of a study at a glance [Figure 3].[3] Figure 3 Pictorial representations of data