3.0  Changes in R broke the package

2.16 fix to cummeta.summaries for NULL names (from BDR)

2.15 fix to MH relative risk variance, from Ted Lystig

     forestplot() now allows the box size to be controlled explicitly

2.14 xticks= argument to forestplot()
     update format of saved data for R 2.4.0

2.13 Add forestplot() for more complicated grid-based plots

2.12 meta.MH for relative risks was wrong when exposed and unexposed 
      groups were very different in size

2.11 codoc changes.

2.10 meta.MH was returning the trial-specific odds-ratio standard
     errors, not relative risk standard errors.

2.9  Allow factors as plot labels.

2.8  Coerce data to double precision to avoid integer overflows.

2.7  add more options to meta.colors.  
     don't draw the zero line over the labels.

2.6  Change meta.cum to cummeta, add cummeta.summaries, 
     Change x-axis tick calculation for log scales

2.5  Random effects meta-analyses weren't actually using the DerSimonian &
     Laird estimator but a different moment estimator.
     <NBarrowman@exchange.cheo.on.ca>

     Now uses the options() default for na.action

     Colours can be specified as a vector, which will be recycled.

2.4-1 bug in meta.cum (ripley@stats.ox.ac.uk)

2.4  Cumulative meta-analyses, changed n-1 to n in formula for tau in
     meta.DSL, allowed relative risk in addition to odds ratio.

2.3  CMD check fixes (thanks, Kurt)

2.1  Better color handling. 
     Funnel plots
     meta-analysis of non-binary data

2.0  User-specified confidence level (mostly from Ko-Kang Wang)
     Colors.