New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Short term work plan #5
Comments
This is a good start. A few ideas I would like to consider:
1) Toy problems where we generate data from:
a) 3 parameter model with different degrees of scatter
b) Conflicting data sets from different underlying parameter sets
2) Include the 1, 2 parameter models (simultaneous sampling/comparison of 4
models)
3) Include 3 parameter model where parameters are eps, sig, and Q
4) Sensitivity to error model, data ranges, amount of data, etc (could use
toy problems for this)
…On Monday, February 11, 2019, Owen Madin ***@***.***> wrote:
- Improve model swap acceptance for the 'All' cases
In general, model swap acceptance ratios are lower for 'All' cases than
they are for 'rhol+psat' cases. We should look at using better Q proposals,
model mapping techniques, and swap frequencies to improve these ratios -
they will probably be different for different compounds
- Run 10^7 step RJMC for all compounds in the 'rhol+Psat' case.
Currently completed:
- C2H6 (Bayes Factor = 2.8)
- N2 (Bayes Factor = 2.4)
- O2 (Bayes Factor = 8.0)
Todo:
-
C2H2 (Good data coverage except for SurfTens, only coverage from
0.62Tc-0.7Tc)
-
C2H4 (Good data coverage except for SurfTens, only coverage from
0.41Tc-0.65Tc)
-
Br2 (Good data coverage except for SurfTens, only coverage from
0.47Tc-0.55Tc)
-
Cl2 (Only Psat data available)
-
C2F4 (Good data coverage except for SurfTens, No SurfTens data
available)
-
C2Cl4 (rhol and Psat data only from 0.4Tc-0.65Tc, no SurfTense data
available)
-
F2 (Good Psat data coverage, rhol only from 0.5Tc-0.6Tc, SurfTens only
from 0.45Tc-0.55Tc)
-
Run 10^7 step RJMC for selected compounds in the 'All' case
-
Have completed simulations for C2H6, N2, but unsure if results are
valid (100% model 0 sampling for N2, uncoverged distribution for C2H6)
-
Will run for O2 in the near future
-
Won't be possible to do this for Cl2,C2F4,C2Cl4
-
Will be able to do with modified data ranges for C2H2,C2H4,Br2,F2
This is a relatively short term plan, after completing these tasks and
looking at the distributions/Bayes factors, we can determine how to move
forward. If there are things you think I am missing please let me know.
@ramess101 <https://github.com/ramess101>
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#5>, or mute the thread
<https://github.com/notifications/unsubscribe-auth/AWUvhDXtOhxeILBkSc85-jOwJ-FAFYjFks5vMgAggaJpZM4a1Sop>
.
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
In general, model swap acceptance ratios are lower for 'All' cases than they are for 'rhol+psat' cases. We should look at using better Q proposals, model mapping techniques, and swap frequencies to improve these ratios - they will probably be different for different compounds
Currently completed:
Todo:
C2H2 (Good data coverage except for SurfTens, only coverage from 0.62Tc-0.7Tc)
C2H4 (Good data coverage except for SurfTens, only coverage from 0.41Tc-0.65Tc)
Br2 (Good data coverage except for SurfTens, only coverage from 0.47Tc-0.55Tc)
Cl2 (Only Psat data available)
C2F4 (Good data coverage except for SurfTens, No SurfTens data available)
C2Cl4 (rhol and Psat data only from 0.4Tc-0.65Tc, no SurfTense data available)
F2 (Good Psat data coverage, rhol only from 0.5Tc-0.6Tc, SurfTens only from 0.45Tc-0.55Tc)
Run 10^7 step RJMC for selected compounds in the 'All' case
Have completed simulations for C2H6, N2, but unsure if results are valid (100% model 0 sampling for N2, uncoverged distribution for C2H6)
Will run for O2 in the near future
Won't be possible to do this for Cl2,C2F4,C2Cl4
Will be able to do with modified data ranges for C2H2,C2H4,Br2,F2
This is a relatively short term plan, after completing these tasks and looking at the distributions/Bayes factors, we can determine how to move forward. If there are things you think I am missing please let me know.
@ramess101
The text was updated successfully, but these errors were encountered: