The quantile operator returns the quantiles of the specified array.

Synopsis

quantile( srcArray, q_num [ , attribute [ , dimension_n ... ] ] )

Summary

A q_quantile is a point taken at a specified interval on a sorted data set that divides the data set into q subsets. The quantiles are the data values marking the boundaries between consecutive subsets.

You specify the source array (srcArray) and the number of quantiles (q_num). Optionally, you can specify an attribute and a dimension for grouping. To group by one or more dimensions, you must specify the attribute.

Note the following:

The quantile operator returns q_num+1 values, which correspond to the lower and upper bounds for each subset.
The quantile operator returns the same datatype as the attribute.
The q_num argument must be a positive integer. Otherwise SciDB returns an error.

Inputs

The quantile operator takes the following arguments:

srcArray: A source array with one or more attributes and one or more dimensions.
q_num: The number of quantiles.
attribute: An optional attribute to use for the quantiles. If you don't specify an attribute, SciDB uses the first one.
dimension_n: An optional list of dimensions to group by.

Examples

Calculate the 2-Quantile for a 1-Dimensional Array

To calculate the 2-quantile for a 1-dimensional array, do the following:

Create a 1-dimensional array called quantile_array:

AFL% create array quantile_array <val:int64>[i=0:10];

The output is:

Query was executed successfully.

Put eleven numerical values between 0 and 11 into quantile_array:

AFL% store(build(quantile_array, '[10,3,0,3,4,5,9,11,7,3,3]', true), quantile_array);

The output is:

{i} val
{0} 10
{1} 3
{2} 0
{3} 3
{4} 4
{5} 5
{6} 9
{7} 11
{8} 7
{9} 3
{10} 3

Find the 2-quantile of quantile_array

AFL% quantile(quantile_array,2);

The output is:

{quantile} percentage,val_quantile
{0} 0,0
{1} 0.5,4
{2} 1,11

Remove the quantile_array

AFL% remove(quantile_array);

The output is:

Query was executed successfully

The Group-by-Dimension Parameter

To see/use the group-by-dimension parameter, do the following:

Start with a 5x5 array, with a single, integer attribute:

AFL% create array m5x5<val:int32>[i=0:4; j=0:4];

The output is:

Query was executed successfully.

Initialize the data in the array

AFL% store(build(m5x5, '[[16,13,22,7,13],[11,19,23,21,24],[16,21,15,7,16],[10,19,0,23,23],[12,7,18,7,8]]', true), m5x5);

The output is:

{i,j} val
{0,0} 16
{0,1} 13
{0,2} 22
{0,3} 7
{0,4} 13
{1,0} 11
{1,1} 19
{1,2} 23
{1,3} 21
{1,4} 24
{2,0} 16
{2,1} 21
{2,2} 15
{2,3} 7
{2,4} 16
{3,0} 10
{3,1} 19
{3,2} 0
{3,3} 23
{3,4} 23
{4,0} 12
{4,1} 7
{4,2} 18
{4,3} 7
{4,4} 8

Find the 2-quantile of the array, and then by the first dimension, and then by the second dimension.

AFL% quantile(m5x5,2);

The output is:

{quantile} percentage,val_quantile
{0} 0,0
{1} 0.5,16
{2} 1,24

AFL% quantile(m5x5,2,val,i);

The output is:

{i,quantile} percentage,val_quantile
{0,0} 0,7
{0,1} 0.5,13
{0,2} 1,22
{1,0} 0,11
{1,1} 0.5,21
{1,2} 1,24
{2,0} 0,7
{2,1} 0.5,16
{2,2} 1,21
{3,0} 0,0
{3,1} 0.5,19
{3,2} 1,23
{4,0} 0,7
{4,1} 0.5,8
{4,2} 1,18

AFL% quantile(m5x5,2,val,j);

The output is:

{j,quantile} percentage,val_quantile
{0,0} 0,10
{0,1} 0.5,12
{0,2} 1,16
{1,0} 0,7
{1,1} 0.5,19
{1,2} 1,21
{2,0} 0,0
{2,1} 0.5,18
{2,2} 1,23
{3,0} 0,7
{3,1} 0.5,7
{3,2} 1,23
{4,0} 0,8
{4,1} 0.5,16
{4,2} 1,24

Remove the array

AFL% remove(m5x5);

The output is:

Query was executed successfully