Do i'm need to learn Statistic/Math before using Stan Probabilistic Programming

General
1

Hi! I’m a newbie in here so i don’t know if math is a necessary for me to learn using Stan or not

2

There are no hard rules in life.

4 Likes
3

Math is definitely helpful, but just a little bit of math will get you a long way.

For using Stan through high level-interfaces such as brms, I’d suggest that the bare minimum critical math that you need is:

  • An informal conceptual understanding of what a probability density function is and what a probability mass function is. This is needed to understand what your model is expressing.
  • A notion of what an average is and what quantiles are. This is needed to interpret MCMC output.
  • For generalized linear models, a notion of what a function is and familiarity with the link and inverse link functions associated with your particular model (e.g. log and exponential in a Poisson regression; logit and logistic in a binomial regression, etc).

To code models in raw stan, I would add:

  • Solid familiarity with logarithms
  • A clear notion of how to manipulate probabilities, including conditional probabilities (e.g. multiply probabilities of independent events)

To code certain models efficiently or to thoroughly understand the bits of stan code that you might encounter on these forums or generated by brms, I would further add:

  • Familiarity with matrices and matrix operations (particularly matrix multiplication)
5 Likes
4

Does those math enough for me to doing Data Analyst with Stan ? I scare that maybe i will stuck somewhere because mine bad math background. Thank you

5

But sometimes there are hard consequences.

1 Like
6

@Trong_Tue, getting stuck and unstuck, and stuck and unstuck again, is a fundamental part of learning Bayesian statistics for any person. Welcome to the club.

2 Likes
7

It makes good learning for stubborn people.