Possible Effect of Loss of Information in Reducing Scales of Measurement

In medicine and social sciences, data are often collected in the form of nominal or ordinal scale of measurement. When data are collected in form of interval or ratio scale of measurement they are usually reduced into either nominal or ordinal form for easy analysis,; thus, this work considered the possible effect of loss of information in reducing scales of measurement when data in form of interval or ratio scale of measurement are collapsed into either nominal or ordinal scale of measurement. The Pearson product moment correlation coefficient for continuous variable X (the human body weight) and Y (the corresponding blood pressure) are computed and the 95% confidence interval ( i l & ii l ) are estimated, the data were later collapsed into various CXC table of various sizes and corresponding Tau-Statistics ( ors l ) were also estimated. It is observed that more than 95% of the time the Tau-Statistics ( ors l ) fells within the 95% confidence intervals for the Pearson product moment correlation coefficient ( cs l )i.ethe two coefficients (TauStatistics and correlation coefficient) compared very favourably showing that the analytical results obtained from both reduced and original data were identical.