Back to Percentiles.

What is the difference between

PERCENTILE.INC and PERCENTILE.EXC?
To start with, there is no common approach to
calculating a percentile!
PERCENTILE.INC

In EXCEL The original PERCENTILE function and the new

PERCENTILE.INC function have the same definition.

With those functions, we can ask "where in the data is the

100th percentile?".

Answer: the highest value

However, the traditional statistical definition of the kth percentile
is: the (interpolated) data point below which k% of the data lie

With that definition, we cannot ask "where is the 100th

percentile?"

Why? Well, simply because there is no data point below

which 100% of the data lie. (Alternatively, there are infinite data
points below which 100% of the data lie, namely any value
above the highest value)

PERCENTILE.EXC tries to fit that traditional statistical definition

Example: quartiles

Q1 Q2 Q3

3 7 9 13 17 18 22 27 29 32

estimating the quartile demarcation value is called interpolation

 QUARTILE.INC: this function computes the interpolation on a n-1
basis (where n is the number of observations)

Q1 Q2 Q3

Values 3 7 9 13 17 18 22 27 29 32
Interpolation basis 0 1 2 3 4 5 6 7 8 9

We draw the line for Q1, at 0.25*(n-1) or 0.25*(10-1) = 2.25

We draw the line for Q2, at 0.50*(n-1) or 0.50*(10-1) = 4.5
We draw the line for Q3, at 0.75*(n-1) or 0.75*(10-1) = 6.75
We compute the quartiles through interpolation. So the first
quartile is given by

10

Values 3 7 9 13 17 18 22 27 29 32
Interpolation basis 0 1 2 3 4 5 6 7 8 9

2.25

13 9 Q1 9
Q1 10
3 2 2.25 2
The median is given by:

10 17.5

Values 3 7 9 13 17 18 22 27 29 32
Interpolation basis 0 1 2 3 4 5 6 7 8 9

2.25 4.5
What about the third quartile?

10 17.5 ??

Values 3 7 9 13 17 18 22 27 29 32
Interpolation basis 0 1 2 3 4 5 6 7 8 9

2.25 4.5 6.75

What is the third quartile?

1. 22.50

2. 23.65
72%

3. 24.25

4. 25.75
25%

3%
0%
0

5
.5

.6

.2

.7
22

23

24

25
What about the third quartile?

10 17.5 25.75

Values 3 7 9 13 17 18 22 27 29 32
Interpolation basis 0 1 2 3 4 5 6 7 8 9

2.25 4.5 6.75

27 22 Q3 22
Q3 25.75
76 6.75 6
QUARTILE.EXC
QUARTILE.EXC: this function computes the interpolation on a
n+1 basis (where n is the number of observations)

Q1 Q2 Q3

Values 3 7 9 13 17 18 22 27 29 32
Interpolation 0 1 2 3 4 5 6 7 8 9 10 11
basis

2.75 5.5 8.25

We draw the line for Q1, at 0.25*(n+1) or 0.25*(10+1) = 2.75

We draw the line for Q2, at 0.50*(n+1) or 0.50*(10+1) = 5.5
We draw the line for Q3, at 0.75*(n+1) or 0.75*(10+1) = 8.25
As before, lets compute the quartiles through interpolation. So
the first quartile is given by

8.5

Values 3 7 9 13 17 18 22 27 29 32
Interpolation 0 1 2 3 4 5 6 7 8 9 10 11
basis

2.75

9 7 Q1 7
Q1 8.5
3 2 2.75 2
The median (Q2) is given by:

17.5

Values 3 7 9 13 17 18 22 27 29 32
Interpolation 0 1 2 3 4 5 6 7 8 9 10 11
basis

5.5

18 17 Q 2 17
Q 2 17.5
65 5.5 5
What is the third quartile (Q3)?

8.5 17.5 ??

Values 3 7 9 13 17 18 22 27 29 32
Interpolation 0 1 2 3 4 5 6 7 8 9 10 11
basis

2.75 5.5 8.25

What is the third quartile?

1. 27.25 0%
0%
14%
2. 27.50
86%

3. 28.00

4. 28.75 27.25 27.50 28.00 28.75

What is the third quartile (Q3)?

8.5 17.5 27.5

Values 3 7 9 13 17 18 22 27 29 32
Interpolation 0 1 2 3 4 5 6 7 8 9 10 11
basis

2.75 5.5 8.25

29 27 Q3 27
Q3 27.5
98 8.25 8
Comparing the .INC and .EXC quartiles

Q1 Q2 Q3 Q4
INC 10 17.5 25.75 32
EXC 8.5 17.5 27.5 NA

The 2nd quartile will always be the same between any quartile
methods you use.
The 2ndquartile is effectively the Median so that is locked.
Comparing the .INC and .EXC quartiles

QUARTILE.EXC
excludes the median (middle number) from the population
This function results in quartiles that are a further from the center of the whole
This gives a better estimate of actual population and gives a potentially more
accurate view of what values should be considered outliers

QUARTILE.INC
includes the median (middle number) when determining interpolation
It is handy if you require relatively symmetrical quartiles
However, it shrinks your interquartile range (distance between the upper and
lower quartiles), making it more difficult to identify true statistical outliers