Learning
Community
Open Exchange
Global Masters
InterSystems IRIS Data Platform 2019.3 / Application Development / InterSystems MDX Reference / MDX Functions / PERCENTILE
Previous section   Next section

PERCENTILE

Evaluates a given expression (or the current measure), across all elements of a set, and returns the value that is at a given percentile level.

Returned Type

This function returns a number.

Syntax and Details

PERCENTILE(set_expression, optional_numeric_expression, optional_percentile_value)
Where:
  • set_expression is an expression that evaluates to a set, typically a set of members or tuples.
  • optional_numeric_expression is a numeric-valued expression that the function evaluates for each set element.
    Typically, this expression has the form [MEASURES].[measure_name]
    If you do not specify a numeric expression, the system uses the measure used by the current result cell. For example, this might be the measure used on the 0 axis or the measure specified in the WHERE clause, if any. If the query itself does not specify a measure, the system instead uses %COUNT, which counts records in the fact table.
  • optional_percentile_value is a numeric literal that represents the percentile to find. For example, use 30 to find the 30th percentile, which is the value that is greater than 30 percent of the other values.
    If you omit this argument, the system computes the 50th percentile.
The function evaluates the numeric value for each element of the set and returns the value that is at the given percentile.
Tip:
To instead find a value at the percentile across the lowest-level records, use the %KPI function with the sample plug-in class %DeepSee.PlugIn.Percentile. To see examples, download and set up the samples as described in “Accessing the Samples Shown in This Book,” in the preface; then display the User Portal in the namespace where you downloaded the samples and look at the sample dashboards in the KPIs & Plug-ins folder.

Example

For reference, the following query shows the Patient Count measure for the members of the aged.year level. The ORDER function sorts these members into order by their value of Patient Count so that we can easily compare the later results to this query:
SELECT MEASURES.[%COUNT] ON 0, ORDER(birthd.year.MEMBERS,MEASURES.[%COUNT],BASC) ON 1 FROM patients
 
                             Patient Count
 1 1916                                   1
 2 1921                                   1
 3 1922                                   1
 4 1925                                   1
 5 1941                                   1
 6 1914                                   2
...
82 1967                                  18
83 1969                                  18
84 1973                                  18
85 1978                                  18
86 1979                                  18
87 1981                                  18
88 2002                                  18
89 2009                                  18
90 1968                                  19
91 1998                                  21
92 1991                                  23
93 2003                                  23
94 1977                                  25
Next, the following query shows the 5th percentile value for these members:
SELECT MEASURES.[%COUNT] ON 0, PERCENTILE(birthd.year.MEMBERS,,5) ON 1 FROM patients
 
                             Patient Count
5 Percentile                              1
That is, the 5th percentile consists of birth years that have at most 1 patient.
The following query shows the 95th percentile instead:
SELECT MEASURES.[%COUNT] ON 0, PERCENTILE(birthd.year.MEMBERS,,95) ON 1 FROM patients
 
                             Patient Count
95 Percentile                            18
That is, the 95th percentile consists of birth years that have 18 or fewer patients.
For another example, we use the second argument for PERCENTILE:
SELECT PERCENTILE(birthd.year.MEMBERS,MEASURES.[%COUNT],50) ON 1 FROM patients
  
50 Percentile                           10
For additional, similar examples, see AVG.

See Also

Previous section   Next section