Define a new MethComp function that will do anythiqng we like on any dataset
with the classical structure:

A new MethComp function that will do anything we like on any dataset
with the classical structure (which I contemplate to define as a Meth
object, where meth and item are factors) would look like:

MethComp( data, # Dataset in Meth form i.e with columns:
                # meth, item, (repl), y
          # Model specification:           with aliases for service/convenience
          alpha = TRUE,                      bias = alpha,
           beta = FALSE,                    slope = beta,
            MxI = TRUE,                    matrix = MxI,
            IxR = "repl" %in% names(data), linked = IxR,
          # Transformation to be used for data
          trans = I,
          # Estimation algorithm
           meth = c("REML","MCMC"),
            ... # parameters passed on to MCmcmc or AltReg )

The from this function is a MethComp object, which is a list with components:

  Model: Description of the model used.
  Trans: List of the transformation and the inverse used prior to modelling.
Convert: M x 5 matrix med en rkke per metode og sjler alpha, beta,
         MxI, IxR, Res
   Data: Orginal dataset.
 AltReg: Output from AltReg if this was used:
         A list with components:
   mcmc: Output from MCmcmc if this was used:
         A mcmc.list object, i.e. a list with n.chains elements, each
         of which is a mcmc object.

The following methods will be defined for a MethComp object:

summary - computes all pairwise conversions with prediction SDs
print   - prints the summary in an intelligent way
plot    - makes plots of the conversions between methods.
          optionally also a Bland-Altman plot

- and specifically for MethComp objects that comes from the mcmc estimation:

check - calls either trace or post for assessing mcmc convergence
trace - traces of the parameters to check the mcmc convergence.
post  - posterior densities of parameters.
pairs - two-dimensional posteriors.

o Cleanup the working of BA.plot and BlandAltman --- shoudl be one function.

o Make a name-space and proper methods. DONE

o Convert examples to use the included mcmc.list objects. DONE

o Implement Kristian Linnet's proportional errors method in the Deming function

o Include a check of slope in BA.est and BA.plot. DONE

o Include check of heteroschedasticity in the BA.est BA.plot &cetera. DONE

o Enable a transformation in BA.plot and BA.est. DONE
 
o Give a class to BA.est result so that either a Bland-Altman plot or a
  conversion plot can be produced from the results. DONE

o Put in datasets from: 
                       Name  Factor levels:
  Bland-Altman 1987    PEFR: Wright1, Wright1, Mini1, Mini2 
  Carter.1984        Enzyme: SucHom,SucPel,Alkphos
  Barnett.1969       VitCap: StSkil, StNew, ExpSkil, ExpNew
