Measures of Central Tendency
Master the mean, median, and mode โ when to use each, how outliers affect them, and how skewness shapes their relationship.
What Is Central Tendency?
When someone hands you a dataset with thousands of numbers, the first question is almost always: "What's the typical value?"
Central tendency measures try to answer this question by finding a single number that best represents the "center" of your data. But here's the catch โ there are multiple ways to define "center," and they can give wildly different answers.
The Mean (Average)
The arithmetic mean is what most people think of as "the average." Add up all the values and divide by how many there are.
Test scores: 72, 85, 90, 68, 95
Mean = (72 + 85 + 90 + 68 + 95) / 5 = 410 / 5 = 82
The mean acts like the "balance point" of the data. If you placed weights on a ruler at positions 68, 72, 85, 90, and 95, the ruler would balance at 82.
Strengths of the mean:
- Uses every data point in the calculation
- Mathematically convenient (many formulas build on it)
- Familiar and intuitive
Weakness of the mean:
- Extremely sensitive to outliers
Salaries at a small company (5 employees): 45,000 ยท 55,000 ยท $500,000 (the CEO)
Mean salary = $138,000
Does $138,000 represent the "typical" salary? Absolutely not โ 4 out of 5 employees earn far less. The CEO's salary pulled the mean way up. This is why news headlines about "average income" can be deeply misleading.
The Median (Middle Value)
The median is the middle value when data is sorted from smallest to largest. Half the data falls below, half above.
- Sort the data from smallest to largest
- If odd number of values: the median is the middle one
- If even number of values: the median is the average of the two middle values
Sorted data: 68, 72, 85, 90, 95
Middle value (3rd of 5) โ Median = 85
Sorted data: 68, 72, 85, 90, 95, 98
Middle two values (3rd and 4th of 6) โ Median = (85 + 90) / 2 = 87.5
Now let's revisit the salary example:
Sorted salaries: 45k, **55k, $500k
Median = $50,000 โ much more representative of the "typical" employee!
The median is robust to outliers because it only cares about position, not magnitude. Whether the CEO earns 5 million, the median stays at $50k.
This is why median household income is usually more meaningful than mean household income. The ultra-wealthy pull the mean way up, but the median better represents "typical."
The Mode (Most Frequent)
The mode is simply the value that appears most often.
Shoe sizes sold today: 8, 9, 9, 10, 9, 11, 8, 9, 10
Mode = 9 (appears 4 times)
This is the only measure of central tendency that works for categorical data. What's the most popular ice cream flavor? The mode tells you.
Special cases:
- No mode: If all values appear equally often (1, 2, 3, 4, 5)
- Bimodal: Two values tie for most frequent (suggests two distinct groups in your data)
- Multimodal: Three or more peaks
A bimodal distribution often means your data contains two different populations mixed together. For instance, heights of adults might be bimodal because men and women have different typical heights.
When Each Measure Fails
Situation | Best Measure | Why |
|---|---|---|
| Symmetric data, no outliers | Mean | Uses all data efficiently |
| Skewed data or outliers present | Median | Resistant to extreme values |
| Categorical data | Mode | Only option that works |
| You want "most common" | Mode | Directly answers this question |
| Income, house prices, wealth | Median | Always skewed right |
| Test scores (roughly symmetric) | Mean | Standard practice |
Skewness: How Shape Affects the Measures
The relationship between mean, median, and mode tells you about the shape of your data distribution:
Symmetric distribution (like a bell curve): Mean โ Median โ Mode โ all three are roughly equal
Right-skewed (positive skew): Mode < Median < Mean โ the mean gets pulled toward the long right tail
Examples: income, house prices, hospital stay lengths
Left-skewed (negative skew): Mean < Median < Mode โ the mean gets pulled toward the long left tail
Examples: age at retirement, exam scores on an easy test
Rule of thumb: If Mean > Median, the data is likely right-skewed. If Mean < Median, likely left-skewed. If they're roughly equal, the data is roughly symmetric.
Beyond the Basics: Other Means
The arithmetic mean isn't the only kind of mean:
Weighted Mean: When some values matter more than others.
Example: Your course grade might weight exams at 60% and homework at 40%. If you scored 80 on exams and 95 on homework:
Weighted mean = (0.6 ร 80 + 0.4 ร 95) / (0.6 + 0.4) = (48 + 38) / 1 = 86
Geometric Mean: Used for growth rates and ratios.
If an investment returns +50% one year and -33% the next: Arithmetic mean return = (+50 - 33)/2 = +8.5%. But you actually broke even! The geometric mean correctly gives โ 0%.
Putting It All Together
Home prices in a neighborhood (in $1000s): 250, 275, 280, 290, 300, 310, 320, 340, 350, 1,200
- Mean: $391,500 โ inflated by the mansion
- Median: $305,000 โ solid representation of typical homes
- Mode: No mode (all unique values)
A real estate agent might quote the mean ("Average home price: $391k!") to make the area sound upscale. A savvy buyer would ask for the median.
Lesson: Always ask "which average?" when someone gives you a single number.