Mean
The Arithmetic Mean
The mean is calculated by summing all values and dividing by the count. It represents the 'balance point' of the data—if your data were weights on a seesaw, the mean is where it balances.
Mean as Balance Point
Think of the mean as a fulcrum. The distances from each point to the mean, weighted by frequency, balance out. Points below the mean 'pull' down; points above 'pull' up.
Sensitivity to Outliers
One extreme value can dramatically shift the mean. Adding 100 to our dataset moved the mean from 6 to 21.7—even though 5 of 6 values are still below 10. This is why mean is called 'non-robust.'
When Mean Works Best
• No extreme outliers
• Interval or ratio data
• You need mathematical properties
• Outliers are present
• Ordinal data
• Open-ended categories
Mean is ideal for symmetric, well-behaved data. It's mathematically tractable (used in variance, regression, etc.). But for skewed data or data with outliers, median is often better.
Interactive Mean Calculator
Add your own numbers to see how the mean changes! Try adding an outlier like 100 to see how sensitive the mean is. Compare how mean and median respond differently to extreme values.
Median
The Middle Value
The median is the middle value when data is sorted. For odd n, it's the center value. For even n, it's the average of the two center values. It divides the distribution in half.
Median with Even Numbers
With an even count (6 values here), there's no single middle value. Take the two central values (6 and 7 at positions 3 and 4) and average them: (6+7)/2 = 6.5.
Outlier Resistance
Median is 'robust'—resistant to outliers. Changing 9 to 100 doesn't affect the median at all. The median only cares about position, not actual values at the extremes.
Median for Skewed Distributions
For skewed distributions like income, median better represents the 'typical' value. High earners inflate the mean, but median shows where most people actually fall.
Mode
Most Frequent Value
The mode is simply the most frequently occurring value. Unlike mean and median, it identifies what's most common rather than a mathematical center. It's the only measure applicable to nominal data.
Bimodal and Multimodal
Data can have one mode (unimodal), two modes (bimodal), or more (multimodal). If all values occur equally often, there's no mode. Bimodal data often suggests two distinct groups.
Mode for Categories
12
27
18
8
Mode is the ONLY measure of central tendency for nominal data. You can't calculate a mean or median of colors, but you can identify which color is most popular.
When Mode Falls Short
Mode = 2, but data is spread 1-10. Mode doesn't represent the 'center.'
Heights: 170.2, 171.5, 168.9...
No exact duplicates = no mode.
Mode has limitations: (1) It may not be near the center of the distribution. (2) Continuous data rarely has exact repeats, making mode meaningless without binning. (3) It ignores most of the data.
Choosing the Right Measure
Symmetric Distributions
In a perfectly symmetric distribution (like the normal distribution), mean, median, and mode are all equal. They all point to the same center. Use any measure—mean is typically preferred for its mathematical properties.
Right-Skewed Distributions
In right-skewed data (income, prices), the long tail pulls the mean toward higher values. Mode < Median < Mean. The median is usually the best measure of 'typical' value here.
Left-Skewed Distributions
In left-skewed data (age at death in developed countries, test scores on easy exams), the tail pulls mean toward lower values. Mean < Median < Mode.
Which Measure to Use?
| Situation | Use | Why |
|---|---|---|
| Symmetric data | Mean | Best math properties |
| Skewed data | Median | Robust to outliers |
| Outliers present | Median | Not affected |
| Categorical data | Mode | Only option |
| Finding 'typical' | Median | Represents middle |
| Further calculations | Mean | Used in formulas |
No single measure is always best. Consider your data type, distribution shape, presence of outliers, and intended use. Report multiple measures when appropriate.
Real-World Examples
Apply your knowledge to practical scenarios
Company Salary Analysis
A company with 100 employees reports: 'Average salary is $85,000.' The CEO earns $2,000,000 while most employees earn between $45,000-$65,000.
Analysis: The mean ($85,000) is misleading because one extreme value (CEO salary) pulls it up. The median ($52,000) better represents the 'typical' employee. This is classic right-skewed data.
Sample Data
{
"mean": 85000,
"median": 52000,
"mode": 50000,
"ceo_salary": 2000000
}Key Insights
- Always ask 'mean or median?' when someone reports an 'average' - especially for income, house prices, or any data with potential outliers.
Class Test Score Comparison
Two classes take the same test. Class A: scores are 70, 72, 74, 76, 78 (mean=74, median=74). Class B: scores are 50, 60, 74, 88, 98 (mean=74, median=74).
Analysis: Both classes have identical mean AND median, but Class B has much more variability. This shows why central tendency alone is insufficient - you need spread measures too.
Key Insights
- Two datasets can have the same center but completely different shapes. Always report spread alongside central tendency.
Mystery of the Double Peak
A shoe store analyzes customer purchases. Mean shoe size = 8.5, but very few customers actually buy size 8.5. The mode reveals two peaks: size 6 and size 11.
Key Insights
- When mean falls in a 'dead zone' with few observations, suspect bimodal data. The mode (or histogram) reveals the true story.
Central Tendency Quiz
Test your understanding • Passing score: 70%