The standard deviation is a statistic that measures the data variability. It is derived from the square root of the distances between each value in the population and the population's mean squared.

Variance = σ^{2} = | Σ(𝑥i-x̄)^{2} |

n |

Sample variance = S^{2} = | Σ(𝑥i-x̄)^{2} |

n-1 |

We use

If you ask a school kid how to measure the variability, he will probably suggest one of the following:

1.

2.

MAD = | Σ_{i}|𝑥_{i}-x̄| |

n |

In other words, why do we use the standard deviation instead of the MAD?

1. In Mathematics it is more elegant to use second power and square root functions than to use the absolute function, hence easier calculation. (squares is continuously differentiable, easy to derive for a minimum or maximum, while absolute is not).

2. Absolute differences grant equal weight to any difference while square differences grant larger weight to a big difference. This may be good when the outlier is correct, a higher weight emphasizes that the data may spread more, but it may be worse when the outlier is incorrect, See example, and the large weight of the incorrect observation influences more the result.

In the following example, 'high' and 'low' are only relative.

MAD - Mean Absolute Deviation, the average of the absolute distances from the average.

Data: 6,6.1,5.8,6.9,5.5,6.2,5.9,6.3.

Average absolute distance from the average:

Standard deviation:

Data: 7,1,6,2,11,3,4,5,2,3,4,5,9,7,5,5,8.

Average absolute distance from the average:

Standard deviation:

Following a visual example of the standard deviation calculation.

The **red circle** in the middle symbolizes the average.

The **radius** represents the absolute difference from the average.

Values: **4, 1, 9, 10**

Average = | 4 + 1 + 9 + 10 | =6 |

4 |

Press the "run" button to run the demonstration!

Value: **4**

Difference (r):**2**

Difference (r):

Value: **1**

Difference (r):**5**

Difference (r):

Value: **9**

Difference (r):**3**

Difference (r):

Value: **10**

Difference (r):**4**

Difference (r):

Value: **2.33** or **9.67**

Standard deviation (r):**3.67**

Standard deviation (r):

Green circle area= | π*2^{2}+π*5^{2}+π*3^{2}+π*4^{2} | =π*3.67^{2} |

4 |

The π doesn't influence the result, it is only for the visualization. If we would use the square shape instead of the cycle we would get the same result.

In the following example, we took a special case when the **standard deviation** and the **mean absolute deviation** are similar.

Even **one outlier in a group of 24** influence dramatically both measurements, the MAD and the standard deviation, but it affected much more the standard deviation.

Data: 5.9,6.4,5.9,6.4,5.8,6.2,5.9,6.3,5.9,6.4,5.9,6.4, 5.8,6.2,5.9,6.3,5.9,6.4,5.9,6.4,5.8,6.2,5.9,6.3

MAD:

Standard deviation (σ):

Data: 5.9,6.4,5.9,6.4,5.8,6.2,5.9,6.3,5.9,6.4,5.9,6.4, 5.8,6.2,5.9,6.3,5.9,6.4,5.9,6.4,5.8,6.2,5.9,8

MAD:

Standard deviation (σ):

Data: 5.9,6.4,5.9,6.4,5.8,6.2,5.9,6.3,5.9,6.4,5.9,6.4, 5.8,6.2,5.9,6.3,5.9,6.4,5.9,6.4,5.8,6.2,5.9,12

MAD:

Standard deviation (σ):