Page 1 of 1

Kernel Density Series Transformation

Posted: Wed May 03, 2017 10:00 pm
by diggetybo
If I have an existing series named x, and I want to store the kernel distribution of x in a series object named x_kernel, is there supported commands for this? The most common Epanechnikov formula would go something like:

Code: Select all

series x_kernel = (3/4)*(1-x^2) ....?


How can I implement the indicator function to truncate everything outside the domain [-1,1] to 0?

Please write the complete equation so I can see how you did it, thank you.

*Edit* On second thought, I guess I would need two series objects. One for the x values, and one for the y values.

Re: Kernel Density Series Transformation

Posted: Fri May 05, 2017 9:48 am
by EViews Glenn
You can store the distribution, but not in series objects as the number of evaluation points for the kernel estimate does not in general correspond to the workfile frequency.

Use Proc/Make Distribution Plot Data... from the menus, or the .distdata proc to store the results in a matrix object.

Re: Kernel Density Series Transformation

Posted: Thu Feb 28, 2019 2:30 am
by paolo.zanghieri
I have a similar problem. I want to rescale a series in the [0,1] interval using the percentiles of the kernel distribution. I guess I can do that with a long series of if, then else commands, but I wonder whether there is a quicker and more sensible way to do it. Thanks