| 1 |
|
| 2 |
//------------- 1st set of posterior draws to find optimal scales for Metropolis. --------------- |
| 3 |
//== number draws for first burn-in ==// //For determining the Metropolis scales only. |
| 4 |
30000 |
| 5 |
|
| 6 |
//------------- 2nd set of posterior draws to to throw away the initial draws. --------------- |
| 7 |
//------------- MCMC burn-in draws once the Metropolis scales are fixed. -------------- |
| 8 |
//== number draws for second burn-in ==// |
| 9 |
10000 |
| 10 |
|
| 11 |
//--------------- 1st set of draws from the posterior ---------------- |
| 12 |
//== number draws to estimate mean and variance ==// |
| 13 |
200000 |
| 14 |
|
| 15 |
//--------------- 2nd set of saved draws from the posterior: saved draws AFTER thinning ---------------- |
| 16 |
//== number draws for modified harmonic mean process ==// |
| 17 |
1000000 |
| 18 |
|
| 19 |
//--------------- Only applied to MHM process: thinnning factor * 2nd set of saved posterior draws ---------------- |
| 20 |
//== thinning factor for modified harmonic mean process ==// |
| 21 |
1 |
| 22 |
|
| 23 |
//------- 1st stage: computing all three tightness factors for Dirichlet. --------- |
| 24 |
//------- 2nd stage: hard-code the second scale factor (in principle, we can do all three). --------- |
| 25 |
//== scale values for Dirichlet distribution ==// |
| 26 |
3 |
| 27 |
|
| 28 |
1.0 1.5 2.0 |
| 29 |
|