search

dstndstn / LearningInference

Notes on statistical inference
GNU General Public License v2.0
9 stars 2 forks source link

different results for same initial parameters #3

Open psarkis opened 9 years ago

psarkis commented 9 years ago

I run the same code by changing the initial slope parameter to m = 4. The first time I get an acceptance fraction: 0.2748 the second time 0.0. I did a plot to see how the slope parameter varies with each iteration, during the second trial, it seems that the code is stuck at m = 4. Does it has to do with how the metropolis-hastings method is being implemented?

drphilmarshall commented 9 years ago

I expect so, yes! Can you free the sampler up by varying its proposal distribution? Perhaps the prior on m is making life difficult?

On Thu, Apr 9, 2015 at 12:26 PM, Paula Sarkis notifications@github.com wrote:

I run the same code by changing the initial slope parameter to m = 4. The first time I get an acceptance fraction: 0.2748 the second time 0.0. I did a plot to see how the slope parameter varies with each iteration, during the second trial, it seems that the code is stuck at m = 4. Does it has to do with how the metropolis-hastings method is being implemented?

— Reply to this email directly or view it on GitHub https://github.com/dstndstn/LearningInference/issues/3.