Naar de inhoud
OneKitly

The Mean Absolute Deviation Lost a Fight It Should Have Won

Gepubliceerd op 13-11-2026 · 3 min leestijd · Dagelijkse calculators

Daniel Okonkwo

Daniel Okonkwo — Front-enddeveloper en techredacteur bij OneKitly

Webprestaties · Bestandsformaten

Getoetst aan 2 bronnen

Profiel bekijken →
Kort gezegd

Ask how far a typical value sits from the average and the direct answer is the mean absolute deviation: add up the distances, ignore the signs, divide by how many there are. For 4, 8, 15, 16, 23, 42 the mean is 18 and that average distance is 9.67. The standard deviation for the same data is 13.49, a full 40% larger, because squaring gives the far-out points extra weight before the square root pulls the answer back. Neither is wrong; they answer different questions, and the one almost everybody is actually asking is the first.

Advertentie

For 4, 8, 15, 16, 23, 42 the average distance from the mean is 9.67 and the standard deviation is 13.49. The first one is what the question actually asked.

Eddington against Fisher, and who was right

The astronomer Arthur Eddington preferred the absolute deviation and said so in print; Ronald Fisher showed in 1920 that for perfectly normal data the standard deviation extracts more information from the same sample, and the profession followed Fisher. Forty years later Tukey pointed out the missing condition: real data is never perfectly normal, and a contamination of a fraction of a per cent by wild values is enough to reverse the ranking. Both men were right about the case they described, and the textbooks kept only one of the two cases.

When the choice actually changes a decision

In stock control, safety margins are usually set at a multiple of the spread, so choosing the larger measure books more stock. In server monitoring, an alert threshold set from the standard deviation fires later after a single bad reading has inflated it, while a threshold from the absolute deviation reacts more evenly. Neither behaviour is universally better, but the two produce different amounts of inventory and different numbers of pages at three in the morning, which makes the choice an engineering decision rather than a matter of taste.

Mean absolute deviation
The two measures side by side
DataMean absolute deviationStandard deviation
2, 4, 6, 8, 102.43.16
4, 8, 15, 16, 23, 429.6713.49
Ratio11.25 to 1.4

Uitgewerkt met onze eigen calculator

Calculator gemiddelde absolute afwijking (MAD)

Gegeven

Getallen (door komma's gescheiden)
2, 4, 6, 8, 10

Resultaat

Gemiddelde
6
Gemiddelde absolute afwijking
2,4

Deze cijfers komen uit de calculator hieronder, ze zijn niet met de hand ingetypt — ze worden opnieuw berekend telkens als de tool verandert.

Reken het door met je eigen cijfers →
Advertentie

Veelgestelde vragen

Is there a fixed conversion between the two?
Only for a normal distribution, where the standard deviation is 1.2533 times the mean absolute deviation — the square root of pi over two. Any other shape breaks the ratio: the six-number example here lands at 1.396 because it has one value far out on the right. Seeing a ratio much above 1.25 is itself a signal that the data has a heavy tail, which makes the comparison of the two a free diagnostic rather than a redundancy.
How is this different from the median absolute deviation?
The one described here averages the distances from the mean; the median absolute deviation takes the median of the distances from the median, which is far more robust again. Both are abbreviated MAD, which causes real confusion in papers and in code. The median version ignores up to half the data being corrupted before it moves at all, at the cost of using very little of the information in clean data — the same trade the median itself makes against the mean.

Artikelen die je misschien interesseren

Alle gidsen →

Vergelijkbare tools

Bronnen

Een fout in dit artikel gezien?