Percentiles – Quick Introduction & Examples

The nth percentile is the value that separates
the lowest n% of values from the other values.

Example: the 10th percentile for body weight is 60 kilos. This means that 10% of all people weigh less than 60 kilos and 90% of people weigh more.

Percentiles - Simple Example

Some fishermen catch and measure 100 trouts. The data thus obtained are in this Googlesheet, partly shown below.

Percentiles Simple Example

So what's the 10th percentile for the length of these trouts? For our 100 observations, this is super easy. We simply

As shown in the screenshot above, observations 10 and 11 both have a length of 31 centimeters. This is the 10th percentile for length as either Excel or SPSS will readily confirm.

Sadly, things are rarely that simple with real life data. For example, how to find the 15th percentile from N = 141 observations?

In this case, we'd better use one or two simple formulas. We'll demonstrate them in order to find the 15th percentile for length.

Percentiles - Rank Formula

Percentile \(pct\) is the value that has \(Rank_{pct}\) defined as

$$Rank_{pct} = \frac{pct}{100} \cdot (N + 1)$$


So the 15th percentile for 100 observations is the observation with rank

$$Rank_{15} = \frac{15}{100} \cdot (100 + 1) = 15.15$$

Sadly, there is no observation with rank 15.15. So we look at the nearest ranks, 15 and 16 in our Googlesheet.

Percentile Non Integer Ranks

Note that

If both values would have been equal -as between ranks 10 and 11, both 31 centimeters- we would have reported this value. However, the 15th percentile is some value between 31 centimeters (rank 15) and 32 centimeters (rank 16).

If may be tempting to simply report the average, 31.5 centimeters. However, 15.15 is closer to rank 15 than rank 16. This is usually taken into account by linear interpolation.

Percentiles - Interpolation Formula

For non integer ranks, exact percentiles are usually computed with

$$Pct = X_{tr} + (X_{tr + 1} - X_{tr}) \cdot ({r - tr})$$


For our example, this results in

$$Pct = 31 + (32 - 31) \cdot ({15.15 - 15}) = 31.15$$

Our Googlesheet shows how to implement this formula and its outcome.

Percentiles Interpolation Formula

Note that we replicated this outcome with the built-in function for percentiles, which is =PERCENTILE.EXC(B2:B101,0.15) in Googlesheets as well as Excel. As we'll see in a minute, SPSS yields the same outcome.


You may have noticed that Excel and Googlesheets contain 2 different percentile formulas:

So which one is best?

My personal opinion is that PERCENTILE.EXC makes more sense given our definition: the nth percentile is the value that separates
the lowest n% of values from the other values.
This implies that the zeroeth percentile would be the value that separates the lowest 0% (?!?!) of all values from the others.

This -and therefore PERCENTILE.INC- doesn't make a lot of sense to me. But if you disagree, I'll be happy to hear from you.

Percentiles in SPSS

SPSS users may first download and open trout.sav. Now, the simplest way to find percentiles is from Analyze SPSS Menu Arrow Descriptive statistics SPSS Menu Arrow Frequencies and fill out the dialogs as shown below.

Percentiles In SPSS Frequencies

A much faster option is to use SPSS syntax like the one shown below.

*Find percentiles 5, 10 and 15 for length.

frequencies length
/percentiles 5 10 15.

Completing these steps confirms once more 31.15 centimeters as the 15th percentile for the lengths of our trouts.

Percentiles In SPSS Output

Quartiles, Median & Boxplots

The percentiles that are most often reported are

These percentiles are often reported in boxplots such as the one shown below.

Boxplot Example With Interpretation

Percentiles - Conceptual Issues

Last but not least, I'd like to point out 2 conceptual issues with percentiles that are mentioned by few text books.

First off, in case of ties, percentiles may not exactly separate the lowest n% of observations from the others. Regarding our first example,

Note that there is no single value here that exactly separates the lowest 10% from all other observations.

The second conceptual issue is the opposite: in some cases, an infinite number of values exactly separate the lowest n% of values. This holds for our second example, which came up with a rank of 15.15.

Remember that ranks 15 and 16 corresponded to 31 and 32 centimeters. Our interpolation formula came up with 15.15 centimeters but

Fortunately, these conceptual issues rarely plague real-world data analysis.

Right, so that'll do. If you've any questions or remarks, please throw me a comment below. Other than that,

Thanks for reading!

Tell us what you think!

*Required field. Your comment will show up after approval from a moderator.