Osborne's remarks on Laura Flint Overstreet's paper on
PLFA profiles:

2nd paragraph, 4th line - "The 1st PC [] accounted for 46.6% 
of the [total sample] variance in the multivariate fatty acid
response, ..."

2nd paragraph, lines 7 and 8 - Figures have different numbers (1.1 or 3.1?)

3rd paragraph, line 2 - "This is confirmed by analysis of variance from the 
[factorial effects] model, where the treatment sum of squares was 62.5%
of the total sum of squares for PC1.

3rd paragraph, line 4 - I wouldn't call R-square "goodness of fit"

3rd paragraph, line 5 - effects [explain] 62.5% of the observed variation 
[in] the first PC.

3rd paragraph, line 5-6 - "This indicates that treatment effects explain 
much more variation along the second PC.  In the biplot, Treatment 1 can
be seen ..."

4th paragraph - Generally, I don't follow the breakdown of variation
where you multiple r^2 and proportion of total sample variance explained
in this paragraph.  That is, I don't understand why the product 
.625 X .466 = .291 is interesting or useful.

5th paragrph - I'm reluctant to talk about the main effects of INPUT
in the presence of an INPUT-by-TILLAGE interaction.  The interaction would
seem to indicate that the effect of INPUT is different for one level of
TILLAGE than for another, and that it might not make sense to average
INPUT effects across different levels of TILLAGE.

6th paragraph - 2nd line [total sample variance] instead of [cumulative error]