# mean vs median geometric interpretation?

I'm looking at this picture on wikipedia, comparing the median and mean of an arbitrary distribution. But what does it mean exactly? From the figure, it looks like the mean is the center of mass of the distribution (in the sense that you could balance the distribution on that point). However, since the median has $$50\%$$ to its left and right, it seems a more reasonable candidate for the center of mass. Which one is it? And what is the geometric interpretation of the other metric? • In a symmetric distribution, (normal, for example) the median is equal to the mean. Both measures are estimates of central tendency, but the mean is morr affected by extreme values and outliers than the median is. One can think of the mean as a "balancing point" ; if our distribution was an object, we could balance it on a fulcrum at the mean – WaveX Nov 24 '18 at 18:18
• @WaveX: but that's what i find weird. in the picture, it seems that the median has 50% of the mass on its left and 50% on its right. So, why does the object balance on the mean instead? – blue_note Nov 24 '18 at 18:20
• Perhaps this link may be able to explain it a little more – WaveX Nov 24 '18 at 18:56

## 3 Answers

The two sets $$\{1,2,3\}$$ and $$\{1,2,100\}$$ have the same median. But the second set's center of mass is at $$34$$.

• So, in the picture, is the histogram balanced on the mean? How does that balance, since more than 50% of the area is on the left? – blue_note Nov 24 '18 at 18:14
• @blue_note Because some of the area is farther from the fulcrum. One little guy 100 feet away from the fulcrum imparts more torque than a big guy only one foot away. – B. Goddard Nov 24 '18 at 18:56

The median is the point that splits a set into two equal subsets. The mean is the average of all elements in the set. Both are measures of center, but the mean is more sensitive to the extreme values (i.e., outliers).

The image is not clear. The prob dist has 50% of its values by quantity to the left of the median, and 50% to the right. Each value is put in ascending order and is counted once in computing the median, no matter its value. The dist has 50% of its mass/"area under the curve" to the left of the mean. This is a quick computation using the formula for expected value (technically the discrete and continuous cases are slightly different but the idea is the same).