You can use any distribution for those parameters so long as that distribution’s domain is positive numbers only. I happen to like the half student t because it has a fat tail as compared to say a gamma distribution.

As for deciding the parameters of the prior distribution, this is the biggest complaint about Bayesian models. Eventually you need to pull something out of a hat, whether that is based on experience or just a random guess. My only advice is to try to be reasonable with the parameters that you choose. If you don’t want the burden choose flat or half-flat distributions then you aren’t implicitly making choices.

]]>Nice to meet you. Recently I also researching on PYMC package.

I’m quite interesting to use this method to predict/simulation football results.

but when I read the script in this article, I cant very understand this two lines:

”’

sd_att = pm.HalfStudentT(‘sd_att’, nu=3, sd=2.5)

sd_def = pm.HalfStudentT(‘sd_def’, nu=3, sd=2.5)

”’

1. Why choose HalfStudentT distribution to get sd_att&sd_def?

2. How to decide nu & sd inside the function?

Thanks for you read my question and if available please give me some feedback or information. Thank you very much! ]]>

iex needs API key.

use yahoo as of 2021.

web.DataReader(‘SPY’, ‘yahoo’, dt.datetime(2000,1,1),dt.datetime.now())

thanks! ]]>

I found the data, link in the 3rd part of the tutorial goes to github:

https://github.com/Rblivingstone/visualization

thx