Timeline for How to estimate probability density function (pdf) from empirical cumulative distribution function (ecdf)?
Current License: CC BY-SA 4.0
6 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Sep 19, 2018 at 4:55 | vote | accept | Lei Huang | ||
| Sep 18, 2018 at 21:56 | answer | added | BruceET | timeline score: 7 | |
| Sep 18, 2018 at 20:03 | comment | added | Lei Huang | @Sycorax Thanks. Are there some standard or common kernel methods for this purpose? | |
| Sep 18, 2018 at 18:57 | comment | added | jld | you can estimate the pdf via the empirical pdf which can be arrived at as the Radon-Nikodym derivative of the ecdf with respect to the counting measure, but that's just a fancy way of counting the proportion of data points with each unique value and if you want an estimate that's absolutely continuous w.r.t. the Lebesgue measure then you'll need to do something else (such as what Sycorax suggested) | |
| Sep 18, 2018 at 18:41 | comment | added | Sycorax♦ | The ecdf is a step function, so the "epdf" is just a finite set of spikes. On the other hand, kernel density methods attempt to "smooth" these spikes into a pdf. | |
| Sep 18, 2018 at 18:26 | history | asked | Lei Huang | CC BY-SA 4.0 |