/aggregate/statistics

URL: http://GPUDB_IP_ADDRESS:GPUDB_PORT/aggregate/statistics

Calculates the requested statistics of the given column(s) in a given table.

The available statistics are:
count (number of total objects), mean, stdv (standard deviation), variance, skew, kurtosis, sum, min, max, weighted_average, cardinality (unique count), estimated_cardinality, percentile, and percentile_rank.

Estimated cardinality is calculated by using the hyperloglog approximation technique.

Percentiles and percentile ranks are approximate and are calculated using the t-digest algorithm. They must include the desired percentile/percentile_rank. To compute multiple percentiles each value must be specified separately (i.e. 'percentile(75.0),percentile(99.0),percentile_rank(1234.56),percentile_rank(-5)').

A second, comma-separated value can be added to the percentile statistic to calculate percentile resolution, e.g., a 50th percentile with 200 resolution would be 'percentile(50,200)'.

The weighted average statistic requires a weight column to be specified in weight_column_name. The weighted average is then defined as the sum of the products of input parameter column_name times the weight_column_name values divided by the sum of the weight_column_name values.

Additional columns can be used in the calculation of statistics via additional_column_names. Values in these columns will be included in the overall aggregate calculation--individual aggregates will not be calculated per additional column. For instance, requesting the count & mean of input parameter column_name x and additional_column_names y & z, where x holds the numbers 1-10, y holds 11-20, and z holds 21-30, would return the total number of x, y, & z values (30), and the single average value across all x, y, & z values (15.5).

The response includes a list of key/value pairs of each statistic requested and its corresponding value.

Input Parameter Description

NameTypeDescription
table_namestringName of the table on which the statistics operation will be performed, in [schema_name.]table_name format, using standard name resolution rules.
column_namestringName of the primary column for which the statistics are to be calculated.
statsstring

Comma separated list of the statistics to calculate, e.g. "sum,mean".

Supported ValuesDescription
countNumber of objects (independent of the given column(s)).
meanArithmetic mean (average), equivalent to sum/count.
stdvSample standard deviation (denominator is count-1).
varianceUnbiased sample variance (denominator is count-1).
skewSkewness (third standardized moment).
kurtosisKurtosis (fourth standardized moment).
sumSum of all values in the column(s).
minMinimum value of the column(s).
maxMaximum value of the column(s).
weighted_averageWeighted arithmetic mean (using the option weight_column_name as the weighting column).
cardinalityNumber of unique values in the column(s).
estimated_cardinalityEstimate (via hyperloglog technique) of the number of unique values in the column(s).
percentileEstimate (via t-digest) of the given percentile of the column(s) (percentile(50.0) will be an approximation of the median). Add a second, comma-separated value to calculate percentile resolution, e.g., 'percentile(75,150)'
percentile_rankEstimate (via t-digest) of the percentile rank of the given value in the column(s) (if the given value is the median of the column(s), percentile_rank(<median>) will return approximately 50.0).
optionsmap of string to strings

Optional parameters. The default value is an empty map ( {} ).

Supported Parameters (keys)Parameter Description
additional_column_namesA list of comma separated column names over which statistics can be accumulated along with the primary column. All columns listed and input parameter column_name must be of the same type. Must not include the column specified in input parameter column_name and no column can be listed twice.
weight_column_nameName of column used as weighting attribute for the weighted average statistic.

Output Parameter Description

The GPUdb server embeds the endpoint response inside a standard response structure which contains status information and the actual response to the query. Here is a description of the various fields of the wrapper:

NameTypeDescription
statusString'OK' or 'ERROR'
messageStringEmpty if success or an error message
data_typeString'aggregate_statistics_request' or 'none' in case of an error
dataStringEmpty string
data_strJSON or String

This embedded JSON represents the result of the /aggregate/statistics endpoint:

NameTypeDescription
statsmap of string to doubles(statistic name, double value) pairs of the requested statistics, including the total count by default.
infomap of string to stringsAdditional information.

Empty string in case of an error.