Pythagorean Means: Three Ways to Average#

The arithmetic mean is not the only way to compute an average. There are three classical types of averages, known as the Pythagorean means:

  1. Arithmetic Mean

  2. Geometric Mean

  3. Harmonic Mean

Each is appropriate for a different type of data. Choosing the wrong mean can lead to incorrect conclusions.

1. Geometric Mean#

\[ GM = \left( \prod_{i=1}^{n} x_i \right)^{\frac{1}{n}} \]

The geometric mean multiplies all values together and then takes the \(n\)-th root. It is used when values change multiplicatively rather than additively.

Common uses:

  • Growth rates

  • Investment returns

  • Population growth

  • Ratios

Why Use Geometric Mean (NOT Arithmetic Mean) for Growth?#

Growth compounds over time, so we must multiply changes instead of adding them.

Example (Investment Growth): Suppose an investment changes by:

  • +10%

  • +20%

  • −5%

We must convert percentages into multipliers (growth factors):

  • +10% → 1.10

  • +20% → 1.20

  • −5% → 0.95

What Happens if We Use Arithmetic Mean?

\[ \text{Arithmetic Mean} = \frac{10 + 20 - 5}{3} = 8.3\% \]

This is incorrect because it ignores compounding.

Apply Geometric Mean (Correct Method)

Multiply growth factors:

\[ 1.10 \times 1.20 \times 0.95 = 1.254 \]

Now, take the cube root (since there are 3 periods):

\[ GM = (1.254)^{1/3} \]
\[ GM \approx 1.078 \]

Convert back to percentage growth:

\[ \text{Average Growth Rate} = 1.078 - 1 = 0.078 = 7.8\% \]

The geometric mean is smaller because it correctly accounts for compounding effects.

Key Idea: When values multiply over time, always use the geometric mean.

Remember:

  • It is less sensitive to extreme large values than the arithmetic mean

  • All values must be positive

  • It is appropriate when data represents relative change (like growth rates or ratios)

Harmonic Mean#

The harmonic mean is another measure of central tendency, but it is designed specifically for averaging rates, ratios, and “per-unit” quantities.. Formula:

\[ HM = \frac{n}{\sum_{i=1}^{n} \frac{1}{x_i}} \]

Where:

  • \(n\) = number of observations

  • \(x_i\) = each value

When Should We Use the Harmonic Mean?

Use the harmonic mean when averaging:

  • Speeds

  • Rates (e.g., km/hour, requests/second)

  • Ratios

  • “Per unit” quantities

Why? Because when dealing with rates, time or denominator matters more than magnitude. Slower rates dominate total time, so they should influence the average more strongly.

Example 1: Average Speed: Imagine you drive:

  • First half of the distance at 60 km/h

  • Second half of the distance at 40 km/h

What is your average speed? Most students instinctively compute:

\[ \text{Arithmetic Mean} = \frac{60 + 40}{2} = 50 \]

But that’s incorrect. Why? Because speed is distance divided by time (speed = distance ÷ time). You spend more time traveling at 40 km/h (the slower speed) than at 60 km/h. Slower speeds should affect the average more (dominate total time). The harmonic mean correctly accounts for this:

\[ HM = \frac{2}{\frac{1}{60} + \frac{1}{40}} \]

This gives:

\[ HM = 48 \text{ km/h} \]

Remember: When averaging speeds over equal distances:

  • Slower speeds dominate total time

  • The harmonic mean gives more weight to smaller values

Example 2: Server Response Time Suppose two servers process requests at:

  • Server A: 100 requests/sec

  • Server B: 10 requests/sec

Even if workload is evenly distributed, the slower server becomes the bottleneck. The harmonic mean reflects this by penalizing smaller rates more heavily.

Why Not Arithmetic Mean?#

The arithmetic mean assumes “Quantities combine additively”. But rates combine through time accumulation, not addition. That’s the difference.

Important Properties:

  • Gives more weight to smaller values

  • Always less than or equal to the arithmetic mean

  • Undefined if any value equals zero

  • Only valid for positive numbers

Remember: Use the harmonic mean when averaging anything that is measured per unit (per hour, per second, per distance, etc.).

Big Picture Comparison#

Mean Type

Best For

Sensitive To

Arithmetic Mean

Additive quantities

Large values

Geometric Mean

Multiplicative growth processes

Relative scale

Harmonic Mean

Rates and ratios

Small values