Instruction for developer

• Before pushing make sure that works the following test can be run:
  1. test/mixedeffect/test_NIGMixedEffect.R

TODO

structure

• Add a file that automatically run all test files that needs to be run be fore pushing to bitbucket.

Methods

• Add joint sampling of Gaussian components?
• Automatic switching between distributions based on TV distance
• Add support for weighted subsampling
• Add support for individual nu

Interface

• Add reasonable starting values
• Add checks of input dimensions to all functions
• Add checks of initial values in estimation

TO DO -- OA

• shape parameter is only 0.5 currently - future work
• all the params of gig can be estimated - future work

NOTES -- OA

• model with no random effects
• add tdistr for random effects to summary, plot etc.
• make model specifications flexible, e.g. Normal, normal, Gaussian are all acceptable
• allow different locs in predict than those from the observed data
• continue with predict.derivative!!!
• setseed_ME - explanations to be added
• put examples
• simulate functions - remove?? see LDMod in simulate function!
• pSubsample2 seems un-used
• consider ~ -1 + ... in fixed input of ngme
• add mean or median as an option to the predict.ngme function
• check calculation of coverages and width in predict.ngme
• plot for excursions
• put a note that if nu reaches the limit use normal
• newdata in predict
• fit.init is expected to be gaussian atm
• controls.init vs fit.init
• process nu never becomes 0.001 for normal nig normal model
• warning nu = 100 atm, not for 99.99
• get rid of the for loop at the end of predict.ngme
• ylabs in plot.ngme
• add forecasting and prediction by fine interval by locs.pred
• rename columns and rows of fisher_est in ngme.fisher