NEWS for package moult 

moult 2.3.0

- Added bootstrap confidence intervals (confint.moult function), estimates of covariance matrix and standard errors.
- Added dfbeta.moult function to check influence of individual observations.
- Updated predict.moult function.
- Components for likelihood function are calculated once and components to return are selected depending on type. 
- Fixed format error in help files.

moult 2.2.1

- Fixed error in as.vector(data.frame).

moult 2.2.0

- Fixed error in the calculation of SE(SD in start of moult). Thanks to Les Underhill and Tanya Scott for picking this up. 

- Moult scores in half steps can be converted to proportion of feather mass grown, e.g. "5 5 5 5 4.5 2 0 0 0" (ms2pfmg).
 

Changes in version 2.1.0

- moult_alternative provides alternative parameterization, with halfway date instead of start of moult. Still in testing phase. At later stage will be incorporated into moult function.

- Addition of prec parameter to define measurement precision of moult index (proportion of feather mass grown). 

- Bug fix in type 3 likelihood, now more likely to complete optimization.

- New vignette for individual primary analysis.


Changes in version 2.0.0
 
  - Likelihood is calculated differently: not estimated by density but by integrating between y - 0.02 and y + 0.02, where y is the moult score, or PFMG.

  - 'moult' function now allows fixed parameters. 

Changes in version 1.4

  -- Fixed error in predict.moult.

Changes in version 1.3

  -- Bug fix: moult function did not work when added covariates for standard deviation in start date. 
     Now works, but still only for categorical covariates. 

  -- Improved row and column names for covariance matrix of parameter estimates. 

Changes in version 1.2

  -- Added reference to article in Journal of Statistical Software. 

Changes in version 1.1

  -- The 'weaver' data set now contains only adult birds, and only observations 
         from years 1988-2005.

  -- 'formula' in function 'moult' has changed to 'moult.index ~ days | x1 + x2 | y1 + y2 | z1', 
         where the x's are covariates for duration, the y's are covariates for mean start date of 
         moult and the z can be a single categorical covariate for the standard deviation in start 
         of moult. 
  
  -- The above change depends on package 'Formula'. 

  -- 'moult' now has a 'data' argument, similar to 'lm' or 'glm'. 

  -- 'predict.moult' can now be used for predicting mean start dates at different covariate settings.