search

stan-dev / stan

Stan development repository. The master branch contains the current release. The develop branch contains the latest stable development. See the Developer Process Wiki for details.
https://mc-stan.org
BSD 3-Clause "New" or "Revised" License
2.6k stars 370 forks source link

clean up sample vs. draw in output and doc #1801

Open bob-carpenter opened 8 years ago

bob-carpenter commented 8 years ago

Summary:

We want to follow Andrew's terminology and not confuse a sample (1 or more draws) with a draw.

Description:

Just run CmdStan.

Current Output:

From CmdStan:

#    num_samples = 1000 (Default)`

From RStan:

Samples were drawn using NUTS(diag_e) at Mon Mar 14 13:19:38 2016.

Expected Output:

CmdStan: num_draws = 1000 RStan: Sample was drawn using NUTS.

Current Version:

v2.9.0

maverickg commented 8 years ago

I still remember we had this discussion about sample. I thought in most cases, we are referring to multiple chains. So I have to use samples there.

betanalpha commented 8 years ago

I would avoid the phrase "draw" altogether as it has the physical connotation of an exact sample. Instead stick with "transition" or "iteration".

On Mar 14, 2016, at 17:21, Bob Carpenter notifications@github.com wrote:

Summary:

We want to follow Andrew's terminology and not confuse a sample (1 or more draws) with a draw.

Description:

fix in output [may need to go to interfaces for this] fix in argument names (wait for Stan 3) fix in manual Reproducible Steps:

Just run CmdStan.

Current Output:

From CmdStan:

num_samples = 1000 (Default)`

From RStan:

Samples were drawn using NUTS(diag_e) at Mon Mar 14 13:19:38 2016. Expected Output:

CmdStan: num_draws = 1000 RStan: Sample was drawn using NUTS.

Current Version:

v2.9.0

— Reply to this email directly or view it on GitHub.

ariddell commented 8 years ago

I'd support switching from iter to draws (or num_draws if that's what CmdStan is going to use) as the parameter's name in PyStan/RStan 3. It's currently iter.

betanalpha commented 8 years ago

Hmm, actually I might recant slightly as there is a difference between "samples" and "exact samples". This is worth discussing in person (and maybe waiting until after I finish this intro to probability theory and computation draft).

bob-carpenter commented 8 years ago

OK, when you and Andrew hash it out, let me know, and we can update the doc accordingly.