Timeline for FindDistributionParameters fails with custom distribution?
Current License: CC BY-SA 3.0
18 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 5, 2019 at 16:41 | vote | accept | chris | ||
| Apr 13, 2017 at 12:55 | history | edited | CommunityBot |
replaced http://mathematica.stackexchange.com/ with https://mathematica.stackexchange.com/
|
|
| Dec 13, 2016 at 21:57 | comment | added | JimB |
Clearly I don't understand your purpose. Giving mu and sigma to FindDistributionParameters without giving them numbers is essential. What is it about having your second example work when removing ?NumberQ doesn't work for you? (I'm not trying to be sarcastic. I'm just clearly missing something.)
|
|
| Dec 13, 2016 at 21:53 | answer | added | JimB | timeline score: 2 | |
| Dec 13, 2016 at 21:35 | comment | added | chris | @JimBaldwin yes but it defeats the purpose. What if I am in a situation where my PDF cannot be known a priori, before mu and sigma have a numerical value? | |
| Dec 13, 2016 at 15:57 | comment | added | JimB |
Just to be clear about @J.M. 's comment: removing ?NumberQ makes all of your examples work.
|
|
| Dec 13, 2016 at 8:33 | comment | added | corey979 | Related. | |
| Dec 13, 2016 at 8:10 | history | edited | chris | CC BY-SA 3.0 |
added 386 characters in body
|
| Dec 13, 2016 at 7:47 | history | edited | chris | CC BY-SA 3.0 |
added 187 characters in body
|
| Dec 13, 2016 at 7:40 | history | edited | chris | CC BY-SA 3.0 |
added 856 characters in body
|
| Dec 13, 2016 at 7:26 | comment | added | J. M.'s missing motivation♦ |
Somehow, I feel that the NumericQ[] is interfering with any symbolic attempts being done under the hood. In any event, you should have used ProbabilityDistribution[] like this: ProbabilityDistribution[Piecewise[{{Exp[-(x - μ)/σ], x >= μ}}, Exp[-(-x + μ)/σ]], {x, -∞, ∞}, Assumptions -> (μ ∈ Reals && σ > 0), Method -> "Normalize"]; note the specification of assumptions, and that the (not necessarily normalized) PDF expression itself should be the first argument.
|
|
| Dec 13, 2016 at 7:21 | comment | added | chris | Ah! That might be a lead then. The question remains of interest? | |
| Dec 13, 2016 at 7:19 | comment | added | J. M.'s missing motivation♦ |
Well, ProbabilityDistribution[] supports the setting Method -> "Normalize" so that the normalization is done on your behalf.
|
|
| Dec 13, 2016 at 7:18 | comment | added | chris | @J.M. just tried while you commented. Seem to produce the same. The reason I want to have a delayed definition is that in real life my PDF is not normalized so I need to normalize it via numerical integration. | |
| Dec 13, 2016 at 7:17 | history | edited | chris | CC BY-SA 3.0 |
added 484 characters in body
|
| Dec 13, 2016 at 7:16 | comment | added | J. M.'s missing motivation♦ |
Have you tried using ProbabilityDistribution[] instead to express your custom distribution?
|
|
| Dec 13, 2016 at 7:08 | history | edited | chris | CC BY-SA 3.0 |
added 100 characters in body
|
| Dec 13, 2016 at 6:26 | history | asked | chris | CC BY-SA 3.0 |