Prior question

Use of Healpix, camb, CLASS, cosmomc, compilers, etc.
Post Reply
Nishanth Sasankan
Posts: 27
Joined: September 07 2016
Affiliation: university of notre dame

Prior question

Post by Nishanth Sasankan » February 21 2018

Hi ,
When introducing the a new variable and its prior in params_CMB_default.ini, we have the following format.

center, min, max, start width, propose width

this I assume is a flat prior. Why are we entering a Center value when max and min are given? Is there some special reason for inputting this ?

The reason I ask is because when I make the likelihood plot for my new variable , I see my peak is always near the center value I enter.

Any Insight is appreciated.

Thanks
Nishanth

Antony Lewis
Posts: 1364
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: Prior question

Post by Antony Lewis » February 21 2018

Chains have to start somewhere, so it uses centre as a central estimate (and disperses start points about that depending on the start width). The final answer once converged should be independent of it, just convergence will be faster if it's not too far off.

Make sure your propose width is small enough that the parameter is moving, otherwise it will take ages to converge.

Nishanth Sasankan
Posts: 27
Joined: September 07 2016
Affiliation: university of notre dame

Prior question

Post by Nishanth Sasankan » February 21 2018

If my range is between 0 and 6 , is this a good start and propose width

center min max start propose
3.00 0.00 6.00 0.001 0.001


Thanks
Nishanth

Post Reply