1. Spatial plots
-----------------
a. Legend "Background representation on the array" --> "Spatial distribution of local background estimates"
b. Legend "Foreground representation on the array" --> "Spatial distribution of features intensities"
c. Layout of the PNGs (array 1) and links (to arrays 2...n) on the main page is irregular - clean up. Use table environment?
d. For each of the additional PNGs (2...n), link could go not directly to the PNG but to a short HTML wrapper that gives a caption with array name (number as well as full name) at the bottom.

2. Design
---------
Use lattice throughout 

3.) Code duplication and maintainability
----------------------------------------
Please avoid duplication of long blocks of code - these should be separated
into functions. In particular, the methods for NChannelSets and AffyBatch 
have a lot of redundancy.

4.) Rendering of the output
----------------------------
Abstract the output from HTML, writeReport should internally use a
suitable meta-language that then can be easily rendered into HTML and
Latex/PDF.


5.) Boxplots 
------------ 
Have three plots (of same width), one with
the red values, one with the green, one with the log-rations.

6.) Heatmap
-----------
a) Same color scheme as for spatial plots should be used.
b) Use the one from lattice (or Greg's heatmap.2?)
heatmap is in the latticeExtra package, 
and help(dendrogramGrob) has a complete example

b) Put colors on the side that show an experimental factor of interest.

7.) Asssessment
---------------
Develop criteria for identifying "outlier" arrays. This could be done
(a) on the box-plots/intensity distributions (Section 3) and (b) the
clustering that underlies the heatmap (Section 4).

Section 2, spatial plots: we could develop a measure of homogeneity to
flag arrays with gross features (perhaps based on Fourier or Wavelet
spectra). Need to check literature, I would expect that this has been
descrined.

8.) Stratification
------------------
Plots for diagnosing stratification by factors such as print-tip 
(Block), Row, Column, GC content: ecdf plots.

9.) Duplicate correlation
-------------------------
If featureData contains a factor that identifies features containing 
the same reporter sequence, we can check for that (scatterplots, 
standard deviation)