01/12 sd[] <- to sd <- for exchangeable, unequal variances [Henri Stephan]

11/11 included a work around in fitBMAgamma0 and qunatile.fitBMSgamma0 that 
      helps in situations where the bias coef estimates fall near or below 0
      removed date check from quantileForecast.fitBMA functions

10/11 improved data compression
      fixed problems with robustness of pop estimates for precip (Zhi Li)

09/11 fixed problems with low precip (Alex Lenkowski)
      added combine function for forecasing models
      compressed the srft data
      added verifRankHist, pitHist, and verifPlot functions

06/10 fixed problems with subsetting for ensembleData

03/10 fixed variance model for gamma0 (precip)

12/09 fixed lower bound in quantBMAnormal

11/09 fixed gamma0 model to handle more low precip cases

09/09 added startup speed specification to ensembleData 
      fixed verification rank histogram to range from 1 to n+1

09/09 fixed gamma0 model to handle cases where there are no 0 forecasts

06/09 modified estimation in gamma0 and gamma to do true EM 
      updated bias correction strategy to handle 0 forc in gamma and gamma0 

05/09 fixed crps calculation for gamma0 and gamma models
      changed call to cdfBMAgamma0 to cdfBMAgamma in cdf.ensembleBMAgamma

03/09 changed nonunique forecast error exit to warning

01/09 added subsetting (via date) for fitted BMA models
      changed ensembleDates to ensembleValidDates
      added forecastHour and initializationTime attrributes to model fits  
      changed crps to CRPS and added modeified crps; changed MAE to mae
      added verifRankHist function

12/08 modified ensembleData function for efficiency    
      added forecastHour requirement for ensembleData
      trainingRule changed to trainingDays
      added wind speed data modeling
      changed eps to tol in control functions

11/08 added pit function

10/08 fixed brierScore so that it works with one threshold
      fixed help file syntax