Too often, statistics are used “as a drunken man uses a lamp post, more for support than illumination” (Sumner 1992). Experts in the field can tell whether your study really needs statistics; if it does, they can help you to plan the statistical part of your study, for example, to estimate the sample size needed to demonstrate a difference (if it exists) and to choose appropriate statistical methods.

Then, when your study is completed, you will encounter another serious matter: how to present the statistical results. About half of such presentations contain statistical errors (Murray 1991). Here are the most common ones.

Using mean when median is meant

In a descriptive study on back pain in pregnancy, the women were asked to bend over with their arms hanging down. The distance between fingertips and floor was then measured. The result (mean and standard deviation) was reported as

12 ± 14 cm

thus ranging between -2 and 26 cm, suggesting that some of the women must have poked their fingertips a couple of centimeters through the floorboards. This surprising conclusion is the result of reporting asymmetrically distributed (skewed) data by using mean (the average) and standard deviation instead of median (the value midway between the lowest and the highest value) and a percentile range, such as the interquartile range (25th to 75th percentile).