Chandra Photon Counts

[xcthulhu]

So I wrote something that downloads some data from the Harvard Smithsonian and grabs the photon arrival times.

This makes for an example of a naturally occurring Poisson distribution with much lower variance than the other examples.

I also added praw into requirements.txt

[CamDavidsonPilon]

Wow awesome! This looks great -- I'll start playing with the new data today.

In the requirements.txt, you've add six and update-checker. Are these two packages necessary for the data pull/overall book, or recommended? I ask only because I've never encountered them before.

[xcthulhu]

Well, they were installed when I typed pip install praw in my virtual environment. I suspect they are required for the latest version of praw...

[CamDavidsonPilon]

ah ok.

I've locally pulled your pull request and am editing some things now. One thing I am hesitant about is the amount of terminology being introduced: memorylessness, gaussian, poisson process etc. I want to reader to feel confident approaching an example, and not feel rushed. I'm thinking of a way to reconcile the two approaches - TBD.

[xcthulhu]

bump?

[CamDavidsonPilon]

Yea I'm still debating this one. I am thinking of including it in a secondary route: users who "get it" (to be defined later) can continue to Chapter 2, users who are still confused can go through this simpler example. I'm just curious about how to present that / implement something like that.

[yarikoptic]

I am noone to this project so feel free to ignore my 1c: IMHO adding an ad-hoc script for download/installation/patching of some software goes quite off the scope of this project having nothing to do with software deployment -- it might create a precedent, and later a bit of wasted time on maintenance/support of anyone happen to run into problems.
Cheers!

[xcthulhu]

@yarikoptic I'm inclined to agree. It probably would make more sense if there was a separate repo for data set generation. For what it's worth you don't actually have to run the script, since I have also checked in the results.

[xcthulhu]

Well, I'm a bit disappointed this is still in limbo.

I could make a Dockerfile if you want a good software deployment solution.

I still maintain that tweeting isn't an example of a Poisson process, and you're better off looking in the hard sciences for elementary examples as fodder for MCMC.

[alexgarel]

@xcthulhu
Sorry for I'm a noob here, but personally I prefer the SMS example for it's more close to our everyday life. The fact that it may not be a poisson distribution, reinforce the demonstration that the bayesian observer makes assumptions which may not be totally true, but permits computation, and are retained as a fair choice.