Mean Value and Mean Deviation

The most commonly used analysis of a series of measurements $x_i = x_1, \dots, x_n$ is calculating the mean value, the minimum, and the maximum as well as the mean deviation:

Unfortunately it does not seem to be common knowledge that there are a number of issues that will probably lead to misinterpreted results without further invstigation:

  1. Sensitivity wrt outliers:

The mean value is highly sensitive with respect to high or low outliers. If your series of measurements contains a few extremely high or low outliers or it contains some moderately high or low outliers, the mean value may not represent that which you anticipated. Although there is only a fraction of outliers in the entire series of measurements they are able to greatly affect the mean value. To prevent this you should always check outliers using 0.1 and 0.9 quantiles. Another useful method for visually identifying outliers is building a histogram from the series of measurements.

Please note that this also applied to the mean deviation because it is, in fact, the mean value of the deviation which is a series of measurements as well, although it was calculated from the original series.

  1. Confidence:

It is often important to know how good the mean value is, i.e. how close would it be to the mean value if a large or even infinite number of measurements were taken. This is called the confidence. Keep in mind that this does not nullify the effect of outliers.

Feedback is always welcome! If you'd like to get in touch with me concerning the contents of this article, please use Twitter.