Extend getSample to coda about bayesiantools

[TankredO]

6a21143
ca095a3

[florianhartig]

Hi Tankred,

is this tested and fully functional?

[florianhartig]

Unit tests added?

[TankredO]

Tests will be added today.

[stefan-paul]

I get an error if I try to use the getSample function for coda objects.
At least for coda objects created by the runMCMC function (which shouldn't
be different from any other coda mcmc objects).

# Create setup
ll <- function(x) sum(dnorm(x, log = T))
setup <- createBayesianSetup(ll, lower = c(-10), upper = c(10))
settings <- list(iterations = 1000, adapt = F, optimize = F)

# Run sampler
out <- runMCMC(setup, sampler = "Metropolis", settings = settings)

# Get sample
sam <- getSample(out$codaChain)

Produces:

Error in getSample.matrix(as.matrix(sampler), parametersOnly = parametersOnly,  : 
  object 'includesProbabilities' not found 

Defining an boolean object 'includesProbabilities' in the global workspace solves the
issue but I guess it should work without doing so.

For coda::mcmc.list objects I get:

Error in data[[1]] : object of type 'closure' is not subsettable

Also documentation of the new functionality is missing.

[TankredO]

Seems like I introduced this error in 5ea3310. The problem is that the methods have different signatures (which they must not have according to cran) and I didn't realize that. I will sort that out as soon as possible.

[TankredO]

bf7f4da
should be solved. Documentation is another issue, the "old" methods (getSample.Matrix / getSample.MCMCSampler ....) do not have an own description, so i did not know if that is desired.

[stefan-paul]

Yes, it works now (again). Thanks for the fix!
In the description I'd just add that the argument 'sampler' can also
be a coda object.

[TankredO]

a35eee6