*************
varSelectIP Version 0.2-1
*************
-First Built October 28 2013
-Adds a function varSelectOBayeslinear for analysis of linear models using mixtures of g priors, which provides posterior credible sets, if desired
it also produces credible set under the selected model and model uncertainty.
-parallelization has been removed from the function varSelectIP due to the inability to depend on Rmpi
*************
varSelectIP Version 0.1-5
*************
- First built Aug 23 2011
- changes the way that the table of models is initialized and updated.
- Suppose that 4 models are to be kept, then the initial models will be
0001
0010
0011
0100
- Each time a model's BF is computed, this table is updated.
*************
varSelectIP Version 0.1-4
*************
- First built Nov 26 2010
- Next built Jan 23 2011
- makes the following changes (from Luis' comments)
- computes the actual Bayes Factor
- orders the observations when doing probit model (OK)
- set default for covariates.retain (OK)
- explain Value section more clearly (OK)
- reference paper on my website for now (to change to GC website later) (OK)
- example (OK)
*************
varSelectIP Version 0.1-3
*************
- First built Nov 20 2010
- Logs ALL the models that are visited.
*************
varSelectIP Version 0.1-1
*************
- First built Nov 06 2010
- Uses parallel computation if slaves are present
- Uses only mvtnorm method, not importance sampling
*************
varSelectIP Version 0.1-0
*************
- First built Nov 02 2010
- Gives the user the option of using either mvtnorm or importance sampling to
compute the Bayes Factor
*** Nov 02 2010: IPVarSel has been deprecated. We now turn to varSelectIP ***
*************
IPVarSel Version 0.2-1
*************
- First built Sep 22 2010
- Some duplication removed from NextModel() function.
- Checking added to see if proposed.model == current.model
*************
IPVarSel Version 0.2-0
*************
- First built Sep 21 2010
- added functionality for probit model. Extremely slow. Need to
remove code where score for a model has been computed before!
*************
IPVarSel Version 0.1-0
*************
- First built Sep 17 2010
- This is a skeleton package. Has to be tested, timed, profiled and
optimized.
TODO:
(1) A lot of computations are being repeated. Please weed them out
(2) Do not have to update table if the model did not change.
(3) How about storing BF for models that have been visited before?
(4) Compare to S.A??
(5) Save every how many iterations..
(6) Save to output file