Too many digits—the presentation of numerical data

We have all been frustrated reading numbers to too many decimal places, the simplest being digital scales in the outpatient clinic where measurements are probably not accurate to more than 10g although the implication of the weight recorded is that the accuracy is much greater. In an excellent leading article this month Tim Cole takes us back to first principles to discuss this and provide sensible, pragmatic guidelines for the presentation of numerical data. It is interesting and helpful to work through. Remember the difference between decimal places and significant figures. The number of significant figures (digits) is the number of all digits ignoring the decimal point, and ignoring all leading and some trailing zeros. Data should be rounded appropriately—not too much, not too little. Clearly, for example 22.68 (95% confidence interval 7.51–73.67) is more effectively and meaningfully written as 23 (95% confidence interval 7.5–74). The various reporting tools are discussed. Significant figures should be considered rather than just decimal places. The general principle is to use two or three significant digits for effect sizes, and one or two significant digits for measures of variability. There is a helpful summary table included with recommendations given for different scenarios. See page 608 …