Comments (6)
I'm sorry but there is not enough info here. i.e. which yellow region you are talking about and what you think is wrong. Please provide more details including plots to illustrate your question
https://github.com/joshspeagle/dynesty/blob/master/demos/Demo%201%20-%20Overview.ipynb -- check here and in other notebooks for how plots should look like.
from dynesty.
Thank you for your reply and I'm sorry for the delay. I want to ask some questions about dynesty results and plots. I'm using dynesty to do some work about the Bayes factor and posterior estimation.
The question I asked earlier is whether the posterior probability distribution results provided by dynesty are the results after the region colored (yellow) in traceplot at -lnX, which means where the samples are weighted? This is the traceplot:
However results are now found to have significant deviations. The result looks like:
35907it [1:54:23, 5.23it/s, batch: 4 | bound: 28 | nc: 1 | ncall: 1027630 | eff(%): 3.449 | loglstar: -93.740 < -86.208 < -88.348 | logz: -139.041 +/- 0.247 | stop: 0.872]
The truth values only show in x1 and x5. Through dynesty, I cannot obtain the original injected values, and I have also tried adjusting the bounds and samples, but the results remain the same. In runplot, compared to the graphs in your example, the likelihood seems to have an issue. :
I am so sorry to bother you. I'm new to mcmc type calculation, so I wanted to ask what adjustments can be made to ensure that I can obtain the correct results.
from dynesty.
It looks like your problem have multiple modes and likely dynesty is not picking up the right one.
In the left panel of the very first plot you can see multiple modes and the code jumping from one to another.
It is hard to give a universal solution to this problem as if you have a very broad prior with many modes, it's impossible to guarantee that you find one.
If your problem is well posed and you are confident that that the true value of model parameters return very high likelihood you may want to try using more walkers/ restrict prior range to help the mode discovery. But it's really hard to tell without having an idea how the likelihood surface looks like
from dynesty.
Thank you for your reply! It works! I get a nice posterior distribution with more walkers in rwalk method. I'm a bit confused about the traceplots of results.
Is the posterior distribution of the parameters provided by dynesty starting from where I marked? (it mean where the samples are weighted?)
from dynesty.
Glad it worked better.
The posterior distribution is not just truncated at the point you have marked. The posterior is constructed from weighting of all the live-points. The weights for live points below the 'yellow region' will be extremely small.
from dynesty.
Thank you very much for your help! It has been really helpful to me, and I will try more calculations again.
from dynesty.
Related Issues (20)
- live point's likelihood not valid HOT 6
- Documentation incorrect for Pool helper object HOT 1
- Importing Nested Sampling chains from file to plot HOT 2
- Using dynesty with npdim HOT 8
- Questions about dyplot.cornerplot HOT 4
- Questions about DynamicNestedSampler setting HOT 4
- Periodic parameters improvement by rotating them
- Ellipsoid check failed HOT 2
- tqdm example HOT 4
- Discretised prior in Dynesty HOT 7
- Loglstar and logZ values HOT 1
- Importance weight PDF shape HOT 4
- How to access posteriors? HOT 1
- Doubt on sampling methods HOT 6
- Run-to-run instability of logZ HOT 8
- Using weighted samples HOT 3
- blob issues when add_live HOT 4
- log means log_e or log_10? HOT 1
- Pool crashes in foor loop with "OSError: [Errno 24] Too many open files" HOT 2
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from dynesty.