Timeline for Fitting model for two normal distributions in PyMC
Current License: CC BY-SA 3.0
10 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 10, 2021 at 17:36 | comment | added | Daniel | @Cam.Davidson.Pilon dead link? | |
| Feb 22, 2014 at 3:31 | comment | added | Cam.Davidson.Pilon | @AlexCoventry likely this error is a result of using PyMC3. The syntax above is PyMC2 | |
| Feb 21, 2014 at 20:58 | comment | added | Alex Coventry | With a fresh github clone of pymc, this example fails for me with "TypeError: No model on context stack, which is needed to use the Normal('x', 0,1) syntax. Add a 'with model:' block". If I remove the "model =" line, and put everything in a "with Model() as model:" block, I get a "TypeError: __init__() got an unexpected keyword argument 'size'" I'd really appreciate some help to get this working. I'm just learning pymc too, and I've been looking for a working example of representing a mixture model but everything I'm finding on the web is failing for me in one way or another. | |
| Dec 27, 2012 at 23:36 | comment | added | Cam.Davidson.Pilon | The choice of gamma has a mathematical reason. The gamma is the conjugate prior of the precision, see table here | |
| Dec 27, 2012 at 23:34 | comment | added | Cam.Davidson.Pilon | Using a Uniform, as in your original example, implies that you know with absolute certainty that the mean does not exceed some value. This is somewhat pathological. It is better to use a normal, as it allows all real numbers to be considered. | |
| Dec 27, 2012 at 23:31 | comment | added | mat kelcey | Not sure I fully understand the true modelling benefit of saying mean1 & mean2 are Normally distributed instead of Uniform (Same goes really for the precision to be honest, I've been using Gamma since "someone else did"). I've got a lot to learn :) | |
| Dec 27, 2012 at 23:29 | vote | accept | mat kelcey | ||
| Dec 27, 2012 at 23:29 | comment | added | mat kelcey | awesome! this approach to the mixing of the two means is exactly what i was trying to get my head around. | |
| Dec 27, 2012 at 23:23 | comment | added | Cam.Davidson.Pilon | Shameless promotion: I just wrote a blog article about Bayes and pyMC literally 1 minute before you posted this, so I invite you to check it out. The Awesome Power of Bayes - Part 1 | |
| Dec 27, 2012 at 23:19 | history | answered | Cam.Davidson.Pilon | CC BY-SA 3.0 |