Data Dive with Subham
Percentiles, Quartiles, IQR and Box Plot Explained: A Complete Step-by-Step Guide
Statistics becomes much easier when a large collection of numbers can be reduced to a few meaningful measurements. Percentiles, quartiles, the median, interquartile range, and box plots are some of the most useful tools for doing exactly that. They allow us to understand where observations are concentrated, how widely the middle portion of the data is spread, and whether unusually large or small observations deserve special attention.
In this detailed guide, we will work through the following dataset from beginning to end:
2, 5, 6, 7, 15, 15, 35, 43, 52, 65, 77, 88, 105, 199, 208, 1000The dataset is deliberately interesting because it contains an extremely large observation, 1000. That observation makes this a useful example for understanding why the median and interquartile range can sometimes describe a dataset better than the mean and ordinary range.
1. Introduction: Why Percentiles and Quartiles Matter
Imagine that you have a list of examination scores, customer spending values, delivery times, salaries, response times, transaction amounts, or website loading times. Looking at every individual observation can quickly become difficult, particularly when a dataset contains hundreds, thousands, or millions of values.
Descriptive statistics solve this problem by compressing a distribution into useful summaries. A percentile tells us where an observation lies relative to the rest of the dataset. Quartiles divide ordered data into approximately four sections. The median identifies the central position. The interquartile range measures the spread of the middle portion of observations. A box plot then puts these ideas into a compact visual form.
These measurements are especially valuable when data is skewed or contains extreme observations. An average can be strongly influenced by unusually large values, whereas the median and IQR focus on the central structure of the distribution.
The example used throughout this article contains sixteen observations. Most observations are relatively small, but the final value is 1000. That single observation creates a substantial right tail and provides a practical demonstration of outlier detection.
2. Understanding the Dataset
Our original dataset contains sixteen numerical observations:
2, 5, 6, 7, 15, 15, 35, 43, 52, 65, 77, 88, 105, 199, 208, 1000Fortunately, the supplied data is already arranged from smallest to largest. This is important because percentile and quartile calculations depend on the relative position of each observation.
We can label the observations using their positions:
| Position | Value | Position as a fraction of n |
|---|---|---|
| 1 | 2 | 1/16 |
| 2 | 5 | 2/16 |
| 3 | 6 | 3/16 |
| 4 | 7 | 4/16 |
| 5 | 15 | 5/16 |
| 6 | 15 | 6/16 |
| 7 | 35 | 7/16 |
| 8 | 43 | 8/16 |
| 9 | 52 | 9/16 |
| 10 | 65 | 10/16 |
| 11 | 77 | 11/16 |
| 12 | 88 | 12/16 |
| 13 | 105 | 13/16 |
| 14 | 199 | 14/16 |
| 15 | 208 | 15/16 |
| 16 | 1000 | 16/16 |
3. Why Sorting Comes First
Percentiles are positional statistics. Position only has meaning when the observations have been ordered. If the same sixteen values were randomly shuffled, the fourth value in the list would no longer necessarily represent a low percentile.
For example, suppose the data were entered as:
52, 2, 1000, 15, 88, 7, 208, 43, 5, 105, 35, 199, 6, 77, 65, 15The fourth item is 15, but that does not mean 15 is the 25th percentile. The list must first be sorted:
2, 5, 6, 7, 15, 15, 35, 43, 52, 65, 77, 88, 105, 199, 208, 1000After sorting, position-based calculations become meaningful.
4. What Are Percentiles?
A percentile describes a location within an ordered distribution. The p-th percentile is commonly interpreted as a value below which approximately p percent of observations lie, depending on the percentile definition being used.
This wording is important. Percentiles are not always calculated using one universal algorithm. Statistical software packages can use different interpolation rules. Therefore, when precision matters, the percentile convention should be stated.
In this worked example, we follow the positional convention used in the supplied calculation:
- The 25% position is associated with the 4th observation.
- The 50% position is represented by the two middle observations, averaged.
- The 75% position is associated with the 12th observation.
This produces Q1 = 7, Q2 = 47.5, and Q3 = 88.
Why can percentile answers differ between calculators?
Different statistical systems may use different definitions for percentile position. Some methods interpolate between observations rather than selecting a single ranked observation. Others use different formulas for converting a percentile into a fractional index. As a result, two mathematically valid software implementations can occasionally produce slightly different quartiles.
The key lesson is not to mix methods halfway through a calculation. Select a method, apply it consistently, and document the convention when the result is being used for formal analysis.
5. Calculating the 25th Percentile: Q1
The first quartile, commonly called Q1, corresponds to the 25th percentile.
There are sixteen observations:
n = 16Twenty-five percent of sixteen is:
Under the positional method used here, the fourth observation is therefore selected for Q1.
Looking at the sorted dataset:
Position: 1 2 3 4 5 6 7 8
Value: 2 5 6 7 15 15 35 43The fourth value is 7.
This means that, under the chosen positional convention, the first quartile is 7 and approximately one quarter of the ordered observations lie at or below this location.
Notice that Q1 is not necessarily the same as the arithmetic average of the first four values. Quartiles are positional statistics, not simply averages of arbitrary groups of numbers.
6. Calculating the 50th Percentile: Median or Q2
The 50th percentile is the median. It is also known as Q2, because it divides the ordered dataset into its lower and upper halves.
Because there are sixteen observations, there is no single observation exactly in the middle. Instead, there are two central observations.
For an even number of observations, the two middle positions are:
and
n/2 + 1 = 9
Therefore, we inspect positions 8 and 9:
Position 8 = 43
Position 9 = 52The median is their arithmetic mean:
The median is particularly informative for this dataset because of the extreme value 1000. A very large observation does not directly change which two observations occupy the middle positions. This is one reason the median is called a robust measure of central tendency.
7. Calculating the 75th Percentile: Q3
The third quartile, or Q3, corresponds to the 75th percentile.
Using the same positional convention:
The twelfth observation is therefore selected:
Position 12 = 88Q3 marks the upper boundary of the middle portion of the data under this quartile convention. Approximately three quarters of the observations are at or below this location, while approximately one quarter are above it.
8. Understanding Q1, Q2 and Q3 Together
The three quartiles provide a compact description of the distribution:
| Statistic | Percentile | Value | Interpretation |
|---|---|---|---|
| Q1 | 25th percentile | 7 | Lower quartile boundary |
| Q2 | 50th percentile | 47.5 | Median or central location |
| Q3 | 75th percentile | 88 | Upper quartile boundary |
These three statistics split the conceptual distribution into four sections. They do not necessarily contain exactly the same number of observations under every percentile definition, especially when a dataset is small or when interpolation is involved. Their primary purpose is to describe relative position.
What do the four quartile regions mean?
- Below Q1: the lower quarter of the distribution.
- Q1 to Q2: the lower-middle portion.
- Q2 to Q3: the upper-middle portion.
- Above Q3: the upper quarter of the distribution.
9. Calculating the Interquartile Range
The interquartile range, abbreviated IQR, measures the spread of the middle 50 percent of a distribution.
Its formula is:
From our calculations:
Q1 = 7 Q3 = 88Therefore:
The central 50% of the distribution spans a width of 81 units under the quartile convention used in this example.
An important property of IQR is that it does not depend directly on the minimum and maximum observations. That makes it much less sensitive to extreme values than the ordinary range.
Range versus IQR
The ordinary range is:
For this dataset:
Compare that with the IQR of 81. The range is enormous because it includes the extreme observation 1000. The IQR focuses on the middle portion and is therefore much more representative of the typical spread.
10. Understanding Outliers
An outlier is an observation that is unusually distant from the main body of a dataset. Outliers can occur because of measurement errors, data-entry mistakes, rare events, legitimate extreme cases, or unusual underlying conditions.
Importantly, an outlier is not automatically an error. Statistical detection identifies observations that are unusual; it does not prove why they are unusual.
One of the most common methods for identifying potential outliers is the 1.5 × IQR rule.
The method constructs two boundaries, often called the lower and upper fences.
Upper Fence = Q3 + 1.5 × IQR
Any observation outside these boundaries is typically flagged as a potential outlier.
11. Calculating the IQR Fences Step by Step
Lower Fence
We know:
Q1 = 7 IQR = 81Substitute these values into the formula:
First calculate the multiplier:
Then subtract:
Therefore, the lower fence is −114.5.
There are no observations below −114.5 because the minimum observation is 2. Therefore, there are no lower-side outliers under the 1.5 IQR rule.
Upper Fence
Now calculate:
Substituting Q3 = 88 and IQR = 81:
We already know:
Therefore:
This point deserves special attention because the originally supplied text states an upper bound of 212.5. That value does not follow from 88 + 1.5 × 81. The arithmetic gives 209.5.
Fortunately, the conclusion about the extreme observation remains unchanged: both 209.5 and 212.5 are below 1000, and 208 is below 209.5. Therefore, 1000 is the only potential upper outlier under these stated quartiles and the standard 1.5 IQR rule.
| Calculation | Result |
|---|---|
| Q1 | 7 |
| Q3 | 88 |
| IQR | 81 |
| 1.5 × IQR | 121.5 |
| Lower fence | −114.5 |
| Upper fence | 209.5 |
| Potential outlier | 1000 |
12. Understanding the Box Plot
A box plot, also called a box-and-whisker plot, summarizes a distribution using its central location, spread, and potential outliers.
The five basic summary values often associated with a box plot are:
- Minimum
- Q1
- Median
- Q3
- Maximum
However, when the 1.5 IQR rule is used, the whiskers usually do not extend all the way to the raw minimum and maximum if those values are potential outliers. Instead, whiskers extend to the most extreme observations that remain inside the fences, while outliers are shown separately.
The five-number summary in this example
| Measure | Value | Meaning |
|---|---|---|
| Minimum | 2 | Smallest observation |
| Q1 | 7 | First quartile |
| Median | 47.5 | Central value |
| Q3 | 88 | Third quartile |
| Maximum | 1000 | Largest observation |
How to read the box
The box extends from Q1 to Q3. In this example, it therefore extends from 7 to 88. Its width represents the IQR, which is 81.
A line inside the box represents the median. Here, that line is located at 47.5.
How to read the whiskers
Under the standard 1.5 IQR box-plot convention, the upper whisker should normally end at the largest observation that is not beyond the upper fence. Since the corrected upper fence is 209.5, the observation 208 remains within the fence. Therefore, the upper whisker reaches 208, while 1000 is plotted separately as a potential outlier.
Similarly, because the lower fence is −114.5, the smallest observation, 2, remains inside the lower fence. Thus, the lower whisker reaches 2.
13. Interactive HTML Representation of the Box Plot
The following simplified visualization represents the important positions. Because the values range from 2 to 1000, the extreme outlier is visually distant from the central box.
Note: This diagram is a conceptual educational representation rather than a precision-scaled statistical chart. The main purpose is to show the relationship between the box, median, whiskers, and extreme observation.
14. Interpreting the Distribution
Calculating statistics is only half the job. The more valuable skill is interpreting what those statistics tell us about the data.
The median is 47.5, while Q1 is 7 and Q3 is 88. This means the central location is substantially lower than the extreme observation of 1000.
The distance from Q1 to the median is:
The distance from the median to Q3 is:
Interestingly, under these quartile values, the median is exactly halfway between Q1 and Q3:
However, this does not mean that the entire distribution is symmetric. The upper tail continues through 105, 199, 208 and 1000, while the lower side stops at 2.
The presence of 1000 is especially important. It is more than four times the upper quartile:
In practical terms, 1000 is dramatically separated from the central portion of the data.
15. Mean vs Median: Why Extreme Values Matter
Another useful way to understand this dataset is to compare the mean and median.
The arithmetic mean is calculated as:
The sum of the sixteen values is:
2 + 5 + 6 + 7 + 15 + 15 + 35 + 43 + 52 + 65 + 77 + 88 + 105 + 199 + 208 + 1000 = 1922Therefore:
The mean is 120.125, whereas the median is only 47.5.
Why is there such a large difference?
The answer is the extreme value 1000. A mean uses every numerical value directly in the sum. Consequently, a very large observation can pull the average upward.
The median, by contrast, depends on the central positions after sorting. The value 1000 is at position 16 and therefore does not occupy either of the two central positions.
| Statistic | Value | Sensitivity to extreme values |
|---|---|---|
| Mean | 120.125 | Relatively sensitive |
| Median | 47.5 | More robust |
| IQR | 81 | More robust |
| Range | 998 | Highly sensitive |
16. CLI Examples: Calculating the Statistics from the Command Line
Command-line tools are useful when working with data repeatedly or when you want to verify a manual calculation programmatically. Before looking at command output, it is useful to see a small code example.
Code Example Before CLI Output
data = [2, 5, 6, 7, 15, 15, 35, 43, 52, 65, 77, 88, 105, 199, 208, 1000] data = sorted(data) n = len(data) q1 = data[3] median = (data[7] + data[8]) / 2 q3 = data[11] iqr = q3 - q1 lower_fence = q1 - 1.5 * iqr upper_fence = q3 + 1.5 * iqr print("Q1:", q1) print("Median:", median) print("Q3:", q3) print("IQR:", iqr) print("Lower fence:", lower_fence) print("Upper fence:", upper_fence) If this program is saved as quartiles.py, you can run it from a terminal:
python quartiles.pySample CLI Output
Q1: 7 Median: 47.5 Q3: 88 IQR: 81 Lower fence: -114.5 Upper fence: 209.5The output verifies the manually calculated values. Notice particularly that the upper fence is 209.5.
Why does the CLI result show 209.5 instead of 212.5?
Because the formula is Q3 + 1.5 × IQR. With Q3 = 88 and IQR = 81, the calculation is 88 + 121.5 = 209.5. The value 212.5 in the original text is an arithmetic inconsistency.
17. Python Example Using a Percentile Library
For production analysis, it is usually better to use a tested statistical library rather than manually assuming that every software package follows the same percentile convention.
A common approach uses NumPy:
import numpy as np data = np.array([ 2, 5, 6, 7, 15, 15, 35, 43, 52, 65, 77, 88, 105, 199, 208, 1000 ]) q1 = np.percentile(data, 25) q2 = np.percentile(data, 50) q3 = np.percentile(data, 75) iqr = q3 - q1 lower_fence = q1 - 1.5 * iqr upper_fence = q3 + 1.5 * iqr print("Q1:", q1) print("Q2:", q2) print("Q3:", q3) print("IQR:", iqr) print("Lower fence:", lower_fence) print("Upper fence:", upper_fence)Depending on the NumPy version and percentile method selected, software output may differ from the positional convention used in this article. This is not necessarily an error. It demonstrates why specifying the percentile method is important.
Explicit Percentile Method
Modern statistical software can often let you specify a percentile method. An explicit method is preferable when reproducibility matters.
import numpy as np data = np.array([ 2, 5, 6, 7, 15, 15, 35, 43, 52, 65, 77, 88, 105, 199, 208, 1000 ]) q1 = np.percentile(data, 25, method="linear") q2 = np.percentile(data, 50, method="linear") q3 = np.percentile(data, 75, method="linear") print(q1) print(q2) print(q3)The important programming lesson is that percentile calculations are not merely about typing 25, 50, and 75 into a function. The underlying method can affect the answer.
18. JavaScript Example for a Browser-Based Calculator
The same statistical concepts can be implemented in JavaScript. This is useful if you are building an educational webpage or a browser-based data analysis tool.
const data = [ 2, 5, 6, 7, 15, 15, 35, 43, 52, 65, 77, 88, 105, 199, 208, 1000 ]; const sorted = [...data].sort((a, b) => a - b); const q1 = sorted[3]; const median = (sorted[7] + sorted[8]) / 2; const q3 = sorted[11]; const iqr = q3 - q1; const lowerFence = q1 - 1.5 * iqr; const upperFence = q3 + 1.5 * iqr; console.log("Sorted:", sorted); console.log("Q1:", q1); console.log("Median:", median); console.log("Q3:", q3); console.log("IQR:", iqr); console.log("Lower Fence:", lowerFence); console.log("Upper Fence:", upperFence);JavaScript's default array sorting can be surprising because it may perform lexicographic sorting when a comparison function is not supplied. For example, numerical values can be treated as strings. That is why the example explicitly uses:
.sort((a, b) => a - b)This tells JavaScript to compare values numerically.
19. Interactive Learning Checks
Testing yourself immediately after learning a concept can help reinforce the calculation process.
Question 1: What is Q1?
Using the percentile convention in this article, what is the 25th percentile?
There are 16 observations and 25% of 16 is 4. The fourth sorted observation is 7.
Question 2: What is the median?
Which two observations are used to calculate the median?
They occupy positions 8 and 9. Their average is 47.5.
Question 3: What is the IQR?
If Q1 = 7 and Q3 = 88, calculate Q3 − Q1.
IQR = 88 − 7 = 81.
Question 4: Is 1000 a potential outlier?
Compare 1000 with the calculated upper fence of 209.5.
Because 1000 is greater than 209.5, it lies beyond the upper fence and is therefore a potential outlier under the 1.5 IQR rule.
20. Important Notes About Percentile Methods
One of the most important statistical details in this entire topic is that the words "25th percentile" do not guarantee one universally identical numerical calculation.
Different statistical texts and software systems have historically used different definitions. Some methods use a rank such as p(n + 1), some use related formulas, and some interpolate between adjacent observations.
For a large dataset, differences between reasonable percentile methods may be relatively small. For a small dataset, the differences can be more noticeable.
Why is method selection important?
Suppose a percentile position falls between two observations. One method may interpolate between them. Another may select one of the observations. A third method may use a different interpolation weighting. Consequently, the final quartile can differ even though each method is internally consistent.
Therefore, if a report, academic assignment, business dashboard, or scientific analysis requires exact reproducibility, document the percentile definition or the software method used.
The method used in this article
The primary worked calculation follows the convention supplied in the original material:
- Q1 is taken as the fourth observation.
- Q2 is the average of the eighth and ninth observations.
- Q3 is taken as the twelfth observation.
Under that convention:
Q1 = 7 Q2 = 47.5 Q3 = 88 IQR = 81This distinction makes the article both faithful to the original calculation and transparent about the statistical convention.
21. Common Mistakes When Calculating Quartiles and IQR
Mistake 1: Forgetting to sort the data
Percentiles are positional. If the data is not ordered, positions do not correspond to percentile locations.
Mistake 2: Treating Q1 as the average of the first four observations
Depending on the chosen quartile method, Q1 can involve interpolation or a positional rule. It is not automatically the arithmetic average of the first four values.
Mistake 3: Using the wrong middle positions
With sixteen observations, the middle positions are 8 and 9, not 7 and 8 or 9 and 10.
Mistake 4: Forgetting that IQR is Q3 minus Q1
It is not maximum minus minimum. Maximum minus minimum is the ordinary range.
Mistake 5: Calling every extreme value an error
An outlier is a statistical classification, not automatically a data-quality failure. A customer with an unusually large transaction might be completely legitimate. A hospital might have a genuinely rare patient outcome.
Mistake 6: Arithmetic errors in the fences
The original calculation illustrates this perfectly. With Q3 = 88 and IQR = 81:
It is therefore important to calculate intermediate values explicitly.
Mistake 7: Assuming the whisker always reaches the raw maximum
In a conventional 1.5 IQR box plot, an observation beyond the upper fence is displayed separately. The whisker reaches the largest non-outlying observation rather than automatically extending to the raw maximum.
22. Real-World Applications of Percentiles and IQR
Percentiles are used across many fields because relative position is often more informative than a raw value.
Education
Examination results can be described using percentile ranks. A student's percentile position indicates how their result compares with the reference population.
Business Analytics
Businesses can analyze customer spending. The median transaction value may describe a typical customer more effectively than the mean when a small number of customers make exceptionally large purchases.
Finance
Financial datasets often contain unusually large gains or losses. Quartiles and robust spread measures can help analysts understand the central behavior without allowing a few extreme observations to dominate every summary.
Web Performance
Response-time analysis is a classic percentile application. Instead of reporting only average response time, an engineering team might examine the 50th, 90th, 95th, or 99th percentile.
This distinction matters because averages can hide tail behavior. A system may have a good average response time while a meaningful percentage of requests remain slow.
Manufacturing
Measurement distributions can be summarized using quartiles and IQR. Observations beyond expected limits can be investigated as possible process anomalies.
Data Cleaning
IQR-based outlier detection is frequently used as an initial exploratory technique. It can help identify observations that deserve further review. It should not automatically be used as a command to delete data.
23. Complete Calculation Summary
Let us bring the entire calculation together in one place.
Step 1: Start with the sorted dataset
2, 5, 6, 7, 15, 15, 35, 43, 52, 65, 77, 88, 105, 199, 208, 1000Step 2: Count the observations
Step 3: Calculate Q1
Fourth observation = 7.
Q1 = 7
Step 4: Calculate Q2
Middle observations are positions 8 and 9:
43 and 52Median = Q2 = 47.5
Step 5: Calculate Q3
Twelfth observation = 88.
Q3 = 88
Step 6: Calculate IQR
IQR = 81
Step 7: Calculate the lower fence
No observation is below −114.5.
Step 8: Calculate the upper fence
Values above 209.5 are potential upper outliers.
Step 9: Identify the outlier
The observations near the upper end are:
105, 199, 208, 1000Both 199 and 208 are below 209.5. The value 1000 is above 209.5.
Final statistical summary
| Statistic | Answer |
|---|---|
| Minimum | 2 |
| Q1 | 7 |
| Median / Q2 | 47.5 |
| Q3 | 88 |
| IQR | 81 |
| Lower fence | −114.5 |
| Upper fence | 209.5 |
| Maximum | 1000 |
| Potential outlier | 1000 |
24. Frequently Asked Questions
What is the 25th percentile?
The 25th percentile is a positional measure indicating the location below which approximately one quarter of the observations fall. It is commonly called Q1, or the first quartile.
What is the 50th percentile?
The 50th percentile is the median, also called Q2. It divides an ordered dataset into two central halves.
What is the 75th percentile?
The 75th percentile is commonly called Q3, or the third quartile. It marks the upper boundary of the central quartile region.
What is the IQR formula?
The interquartile range is calculated as Q3 minus Q1: IQR = Q3 − Q1.
Why is IQR useful?
IQR measures the spread of the middle portion of a dataset and is generally less affected by extreme observations than the ordinary range.
What is the 1.5 IQR rule?
It defines potential outlier boundaries using Q1 − 1.5 × IQR and Q3 + 1.5 × IQR. Observations outside these boundaries are potential outliers.
Is 1000 definitely a bad data point?
No. It is a statistical potential outlier. Additional context is required to determine whether it is an error, a legitimate rare observation, or something else.
Why is the upper fence 209.5?
Because Q3 is 88 and IQR is 81. Therefore: 88 + 1.5 × 81 = 88 + 121.5 = 209.5. The value 212.5 stated in the original material is an arithmetic error.
Why is 208 not an outlier?
Under the stated quartiles, the upper fence is 209.5. Since 208 is below 209.5, it remains inside the upper fence.
Why does the median equal 47.5?
There are sixteen observations, so the two central observations are the eighth and ninth values. They are 43 and 52. Their average is 47.5.
Can different software give different quartiles?
Yes. Different percentile definitions may use different positional and interpolation rules. For reproducible analysis, specify the percentile method used.
25. Key Takeaways for Learners
- Always sort the data before performing positional percentile calculations.
- Q1 corresponds to the 25th percentile.
- Q2 corresponds to the 50th percentile and is the median.
- Q3 corresponds to the 75th percentile.
- IQR = Q3 − Q1.
- The IQR focuses on the middle portion of the distribution and is more robust to extreme observations than the ordinary range.
- The 1.5 × IQR rule is a standard method for identifying potential outliers.
- An observation flagged as an outlier is not automatically an error.
- Box plots provide a visual summary of quartiles, median, whiskers, and potential outliers.
- Percentile methods can differ, so the method should be documented when exact reproducibility matters.
- In this example, Q1 = 7, median = 47.5, Q3 = 88, and IQR = 81.
- Using the standard 1.5 × IQR rule, the upper fence is 209.5 and 1000 is the only potential upper outlier.
🎯 The entire problem in one line
Sorted data → Q1 = 7 → Median = 47.5 → Q3 = 88 → IQR = 81 → Upper fence = 209.5 → 1000 is a potential outlier.
26. Copy-Friendly Configuration
The following compact configuration can be useful when documenting the analysis in another project or teaching environment.
{ "data": [ 2, 5, 6, 7, 15, 15, 35, 43, 52, 65, 77, 88, 105, 199, 208, 1000 ], "quartiles": { "q1": 7, "q2": 47.5, "q3": 88 }, "iqr": 81, "fences": { "lower": -114.5, "upper": 209.5 }, "potential_outliers": [ 1000 ] }
No comments:
Post a Comment