Timeline for How to define function in OpenBUGS
Current License: CC BY-SA 3.0
18 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 16, 2014 at 11:16 | comment | added | indiajoe | Sure. But when i tried before, stackexchange said I cannot upvote without 15 reputations :( Hence I thought I shall explore some more and come back and accept the answer. | |
| Oct 16, 2014 at 5:18 | comment | added | fileunderwater | @indiajoe Great! Remember to upvote if you find the answer useful. | |
| Oct 15, 2014 at 23:18 | comment | added | indiajoe | Thank you very much for the help. I shall try STAN, I am also now giving a try with PyMC3. | |
| Oct 15, 2014 at 13:15 | comment | added | fileunderwater | @indiajoe Response to your update; I dont think this is possible but I'm not sure. Also see my edited answer. | |
| Oct 15, 2014 at 13:14 | history | edited | fileunderwater | CC BY-SA 3.0 |
addition
|
| Oct 14, 2014 at 11:16 | comment | added | indiajoe | I have added a diagram of model also to clarify. | |
| Oct 13, 2014 at 23:20 | comment | added | indiajoe | I have updated my question by giving the model with two layer equations. Is it clear? | |
| Oct 13, 2014 at 23:01 | comment | added | indiajoe | I am sorry for the confusion. May be I shouldn't have used the term input for t,u and v. They are the parameters I need to estimate. ie, in the example you provided they are the betas. What I was trying to say is evaluation of the function M, which does not have any simple analytical form to write using standard bugs functions has to come from simulation. | |
| Oct 13, 2014 at 22:51 | comment | added | fileunderwater | @indiajoe You need to specify more information in your question - in there you write that t, u and v are input, but here in the comments you write that you want to "...find the t, u and v" using M. | |
| Oct 13, 2014 at 22:45 | history | edited | fileunderwater | CC BY-SA 3.0 |
addition
|
| Oct 13, 2014 at 22:43 | comment | added | indiajoe | I have measurements y . which i need to fit with M to find the t,u and v. It is Hierarchical parameter estimation because my y data is grouped, and some of the parameter in M are group specific. | |
| Oct 13, 2014 at 22:31 | comment | added | fileunderwater | @indiajoe You can always just give the output of M as data (basically giving yhat for each observation), and e.g. estimate sigma_y and evaluate the fit of the model. However, from my perspective, you usually want to estimate effects of t,u and v which you cannot do (from what I know) if you only supply a given output of M. What is the desired "outcome" of your MCMC model (i.e. what effects and model features do you want to evaluate)? | |
| Oct 13, 2014 at 22:25 | history | edited | fileunderwater | CC BY-SA 3.0 |
addition
|
| Oct 13, 2014 at 22:24 | comment | added | indiajoe | For each combination of parameters t,u,v; M is calculated using a complicated computer simulation. (It is basically a radiation transfer code). This is calculated already on a cluster machine for a grid of t,u and v. So I was planning to run MCMC on this grid. I can use some interpolation to find M(t,u,v) for each (t,u,v) in the MC chain. But for this i need to define a function which does that in OpenBUGS. | |
| Oct 13, 2014 at 22:01 | comment | added | fileunderwater | Do you mean that all values of M(t,v,u) are "pre-calculated" as a grid of all combinations of predictor variables? What are your predictors, and I assume that you want to estimate the effects of t, v, & u? What type of functions are hidden in M? Without this info it's hard to say what can and cannot be done in Bugs. | |
| Oct 13, 2014 at 21:54 | comment | added | indiajoe | Thank you for the help. That is exactly what I want to do. But my issue is how to define M(t,u,v). I cannot write it as a function of standard distributions available in BUGS. I need to write a function which reads from a pre-calculated grid of t,u,v. How do I accomplish that in BUGS? | |
| Oct 13, 2014 at 21:51 | history | edited | fileunderwater | CC BY-SA 3.0 |
clarification and generalization
|
| Oct 13, 2014 at 17:45 | history | answered | fileunderwater | CC BY-SA 3.0 |