Minimize the number of functions

[jrnold]

Perhaps a naming scheme like:ssm_(filter|smoother|sim|sample)_([mu][mn]).

• filter: Forward filter
• smoother: smoothers
• sample:

with postfixes

• m: multivariate
• u: univariate
• m: missing-values
• n: non-missingvalues
• c: constant? or time-invariant

[jrnold]

Instead minimize the number of functions. I only need the functions necessary to calculate the marginal likelihood and sample from the posterior.

• filter
• lpdf
  -state conditional mean
• simulation
• sample from posterior

This is the minimal package. The other smothers and filters are not as important for Bayesian methods. I need to get this working for the batch and then sequential methods. Having the other functions is more to debug and maintain.

Keep the existing work in a branch, and then clean out the code in the master.