Python Forum
Get underlying function from Kernel Density Estimation - Printable Version

+- Python Forum (https://python-forum.io)
+-- Forum: Python Coding (https://python-forum.io/Forum-Python-Coding)
+--- Forum: Data Science (https://python-forum.io/Forum-Data-Science)
+--- Thread: Get underlying function from Kernel Density Estimation (/Thread-Get-underlying-function-from-Kernel-Density-Estimation)



Get underlying function from Kernel Density Estimation - jpython - Dec-04-2019

Hi everyone,

There are several libraries that allow us to estimate a probability density function using Kerndel Density Estimation.

My question is how I can see the estimated function, not as a plot but as a formula/equation.

I hope you understand what I mean.

Best regards,
jpython


RE: Get underlying function from Kernel Density Estimation - scidam - Dec-04-2019

There is no simple representation of such formula. You are likely looking for a simple one. You can find explanation about kde-estimations at Wikipedia. So, kde-estimation is a function defined as a sequence of calculations (these calculations are hard to perform manually, but easy to do with help of the computer), it is a computational procedure.

If you want to get a "convenient" formula as an estimation of pdf, you can look at parametric probability density estimations, e.g. fit a normal distribution, or fit a mixture of normal distributions (using mixture of known distributions might be very flexible, if you want to handle non-normally distributed data).


RE: Get underlying function from Kernel Density Estimation - jpython - Dec-05-2019

(Dec-04-2019, 11:51 PM)scidam Wrote: There is no simple representation of such formula. You are likely looking for a simple one. You can find explanation about kde-estimations at Wikipedia. So, kde-estimation is a function defined as a sequence of calculations (these calculations are hard to perform manually, but easy to do with help of the computer), it is a computational procedure.

If you want to get a "convenient" formula as an estimation of pdf, you can look at parametric probability density estimations, e.g. fit a normal distribution, or fit a mixture of normal distributions (using mixture of known distributions might be very flexible, if you want to handle non-normally distributed data).

Hi Scidam,

I am able to use the methods you mentioned, but only to evaluate the the probability density function at a certain point. I woul like the formula in mathematical notation. Is this possible?


RE: Get underlying function from Kernel Density Estimation - jpython - Dec-05-2019

(Dec-05-2019, 08:22 AM)jpython Wrote:
(Dec-04-2019, 11:51 PM)scidam Wrote: There is no simple representation of such formula. You are likely looking for a simple one. You can find explanation about kde-estimations at Wikipedia. So, kde-estimation is a function defined as a sequence of calculations (these calculations are hard to perform manually, but easy to do with help of the computer), it is a computational procedure.

If you want to get a "convenient" formula as an estimation of pdf, you can look at parametric probability density estimations, e.g. fit a normal distribution, or fit a mixture of normal distributions (using mixture of known distributions might be very flexible, if you want to handle non-normally distributed data).

Hi Scidam,

I am able to use the methods you mentioned, but only to evaluate the the probability density function at a certain point. I woul like the formula in mathematical notation. Is this possible?

My apologies, I did not read your post carefully. I did look briefly into parametric probability density estimations but again I was not able to print the underlying function it uses to do the estimation.