##### Version 0.4.0
* Fixed a serious bug that caused incorrect predictions when the supplied block
  order differed from the column order of the input data. This affected models in
  which the block priority was changed manually or determined from the data. In
  previous versions, estimated coefficients could be matched to the wrong
  variables during prediction, which may have affected prediction performance and
  benchmark results. Coefficients are now stored and used in the same order as the
  original input variables.

##### Version 0.3.0
* Options added that can deal with blockwise missing data