CHANGES for BMA 3.17
bic.glm:
- fixed the mechanism for dropping columns when there are more than maxCol
  (reported by Patrick Ball)
- changed documentation to say that maxCol does not include intercept
 (it did not in the past but the documentation was wrong).

CHANGES for BMA 3.16
- replaced leaps() in bicreg() with regsubsets() to allow really.big = TRUE
- removed deprecated \synopsis and updated man files accordingly

CHANGES for BMA 3.15.3.2
- removed call to .First.lib in zzz.R

CHANGES for BMA 3.15.3.1
- man examples faster run time
- @stat --> @u

CHANGES for BMA 3.15.3
- added check for length(dropped) == 0 in bic.glm.data.frame and bic.glm.matrix
  reported by Patrick Ball

CHANGES for BMA 3.15.2
- added drop = F to backsolve in glib.R (reported by Brian Ripley)

CHANGES for BMA 3.15.1
- fixed (1-r2/100) check for correlation in bicreg

CHANGES for BMA 3.15
- added predict functions for bicreg and bic.glm
- explained use of OR.fix=Inf to avoid BIC approximation step in bic.glm

CHANGES for BMA 3.14.1
- changed handling of names for glm and cox output in iBMA for compatibility 
  with R 2.14.0

CHANGES for BMA 3.14
- added drop = FALSE in bic.glm.R (reported by Hana Sevcikova)

CHANGES for BMA 3.13
- error in bicreg.R (reported by Kenneth Lo)

CHANGES for BMA 3.12
- error in iBMA.Rd (help file)

CHANGES for BMA 3.11
- fixed imageplot.bma to handle case where there is only one best model
- fixed glib.matrix to handle Gaussian, identity case

CHANGES for BMA 3.10
- changed on of the examples for bic.surv to match changes made to the pbc data set in the survival package

CHANGES for BMA 3.04-09
- fixed summary.bicreg to handle univariate case
- fixed problems in help files
  For.MC3.REG.Rd, MC3.REG.Rd, glib.Rd and race.Rd

CHANGES for BMA 3.03
- fixed problem with imageplot colors

- fixed overly long R statements in documentation

- change in maintainer

CHANGES for BMA 3.02
- fixed problems in bic.glm that occurred when a two column response was specified for the binomial model, for example "cbind(y,n-y)~."

- fixed problem in bicreg that could cause multiple models having posterior probability of 1 when very large in magnitude bic values occur

- modified imageplot.bma
    - changed color scheme so that variables with positive estimates appear as one color and variables with negative estimates occur as a different color.
    - added options for ordering of variables (see documentation for details).

- modified bicreg to return "mle" as a return parameter (equal to ols)

- modified bicreg to work when only one model is selected

- added methods for iterated BMA: iBMA.glm, iBMA.bicreg and iBMA.surv, together with summary and print functions for these methods, and a new graphic function 'orderplot'.