Statistical Glossary
Mean Values (Comparison):
The numerical example below illustrates basic properties of various descriptive statistics with "mean" in their name, like the arithmetic mean , the trimmed mean , the geometric mean , the harmonic mean , and several power mean . Because the last three statistics are defined only for non-negative values, all examples comprise only positive values.
#
|
Sample
|
Type of the mean
|
|||||
arithmetic
|
|
geometric
|
harmonic
|
power
|
|||
p=2
|
p=8
|
||||||
1
|
1, 1, 1, 1, 1
|
1.00
|
1.00
|
1.00
|
1.00
|
1.00
|
1.00
|
2
|
1, 1, 1, 1, 100
|
20.80
|
1.00
|
2.51
|
1.25
|
44.7
|
81.8
|
3
|
1, 1, 1, 1, 0.01
|
0.80
|
1.00
|
0.40
|
0.05
|
0.89
|
0.97
|
4
|
1, 10, 100, 1000, 10000
|
2222.2
|
370.0
|
100.0
|
4.50
|
4494.6
|
8177.7
|
5
|
1, 2, 3, 4, 5
|
3.00
|
3.00
|
2.61
|
2.19
|
3.32
|
4.18
|
6
|
8, 9, 10, 11, 12
|
10.00
|
10.00
|
9.90
|
9.80
|
10.10
|
10.61
|
Row 1 shows that for a set of equal values all mean values coincide with this value.
Row 2 shows the effect of a single extremely large observation (100) in a sample, e.g. because of a single anomalous measurement. The arithmetic mean is changed from 1.0 to 20, as compared to row 1; the trimmed mean is not affected at all (this illustrates the robustness of the trimmed mean). The harmonic mean has changed only from 1.0 to 1.25. The power mean values are affected most of all, and the power mean with
(from 1 to 44.7).
Row 3 illustrates the effect of a single abnormally small value (0.01) in the sample. The harmonic mean is extremely sensitive to such a small element - it has changed from 1.0 to 0.05.
Row 4 includes 5 consecutive powers of 10 -
Row 5 comprises symmetrically distributed values, with
Row 6 comprises values from row 5 shifted by 7.0 units upwards. This illustrates shift invariance of the arithmetic mean and the trimmed mean (which has changed their values by 7.0 units as compared to row 5). It also demonstrates the lack of shift invariance of other types of means (that changed by a value differing from 7.0).