Kernel density estimation function

This function fits observed data points with a smooth peak function to simulate the true probability distribution curve.

• Syntax

  select kernel_density_estimation(bigint stamp, double value, varchar kernelType)

• Parameters

  Parameter	Description
  stamp	Unix timestamp in seconds.
  value	Observed value at the timestamp.
  kernelType	box: A rectangular (uniform) kernel. epanechnikov: An Epanechnikov curve. gaussian: A Gaussian curve.

• Outputs

  Field	Description
  unixtime	Unix timestamp of the raw data.
  real	Observed value.
  pdf	Probability density value at each data point.

• Example

  • Sample query:

    * | 
    select 
        date_trunc('second', cast(t1[1] as bigint)) as time, t1[2] as real, t1[3] as pdf from (
            select kernel_density_estimation(time, num, 'gaussian') as res from ( 
                select '("__time__" - ("__time__" % 10))' as time, COUNT(*) * 1.0 as num from log group by time order by time)
            ), unnest(res) as t(t1)  limit 1000

  • Result: