Dresden 2020 – wissenschaftliches Programm
Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
AKjDPG: Arbeitskreis junge DPG
AKjDPG 6: Hacky Hour
AKjDPG 6.4: Vortrag
Freitag, 20. März 2020, 10:30–10:45, HSZ 105
Better than histograms: Kernel density estimators and why you should use them — Alexandra Völkel1 and •Simeon Völkel2 — 1Universität Bayreuth, Experimentalphysik VIII, Universitätsstraße 30, 95447 Bayreuth, Germany — 2Universität Bayreuth, Experimentalphysik V, Universitätsstraße 30, 95447 Bayreuth, Germany
We show why everyone who has ever made a histogram should learn about kernel density estimation.
Histograms, as commonly used for estimating probability densities, are far from being optimal. In addition, they require two parameters, bin width and position, to be chosen.
Kernel density estimatiors are an easy to use drop-in replacement for virtually all histograms you ever wanted to draw. They combine superior mathematical properties with an at least as intuitive presentation. Their bandwidth, being the only parameter, can be chosen in an optimal sense automatically and adaptively.
Regarding practical application, we discuss their usage in gnuplot and python.