REMARKS BEGIN HERE:

Page 3, lines 12-13: "... pH values below 8.6 and anionic at higher values."

Page 4, line 14:  "Use of novel chemical means of delivery [has] not been..."

Page 5, lines 7-12: "Our first objectiive was to test the  hypothesis that
the oryzalin molecule in either the neutral [or] anionic form may have 
charged regions that would facilitate adsorption to resins. This would 
be ... Our second objective was to determine ... Our third"

Page 7, line 21: "A second  experiment[] was conducted ..."

Page 8, lines 1-4:  I was unable to understand  the analysis of  data from
these four lines, so  we should probably embellish them to tell  the whole
story.

Page 9, I'm inclined to go with the repeated measures  model, though
the way you've justified the fixed effects model is very good, and
considerably less messy.  The problem with the repeated measures model
is that F-tests may be approximate because of design imbalance.  I've  
attached a little SAS PROC MIXED code that might get you started if you 
wanted to try a split-plot model, with funnel  as thhe whole plot unit.  
Of course the  pvalues for  the between subjects factors will go up a 
little because MS(funnel),  or some  linear combo of variance components 
will be used as the error term.

Page 10, line 7: "... above the pKa of oryzalin. [T]his explains the 
greater ..." 

Page 10, line 11: "so[]lution"

Page 10, ADSORPTION EXPERIMENT section.  Should we say something about the
experiment effect?  When  I compare the columns  corresponding to
concentrations observed in both experiments (1000 and 10000 mg/liter)
I see higher oryzalin adsorption in the 2nd experiment.  Should we consider
an analysis in which experiment is taken as a random effect?

Did you have to transform the % adsorption to fit a general linear
model?  If so, are you just reporting % on the original scale in the
table?  If so, that's fine,  but maybe we should mention it in the
discussion on page 10.

By the way, do you know about the SLICE= option for the LSMEANS statement
in SAS?  It is pretty nifty.  Suppose you wanted  only to test for a
resin effect separately at each of the 8 concXpH combos in the 1st part
of the adsorption expt summarized in Table 2.  (You could do this instead
of making all 3*8=24 pairwise comparison of rows within columns.)  You
can just use this command within PROC GLM:

LSMEANS resin*conc*ph/slice=conc*pH;

In the attached code you could carry out inference for pH effects in
the desorption experiments with single F-ratios (and pvalues) instead
of the 3 pairwise comparisons you currently have in Table 5.

Page 12, line 3, I think you mean Table 5 not Table 3.

Page 12, 1st paragraph.  I just make the observation that as
pH increases, desorption decreases in every case except 
experiment 1, event class 3 (events 21-30).  One other suggestion
is that we consider "experiment" as a random effect.  How else do
we explain the different outcomes across experiments?


Figure 2 (p. 19)
Endpoints of confidence intervals for slopes could be switched.
For example, (.291,.339) might look better than (.339,.291).

Tables 2 (p. 21,22)  Is it possible to report a single minimum significant
difference for any pairwise comparison?  I'd guess it would be of the form 
MSD = q(.05,24,48)*sqrt(MSE/3) where q(.05,24,48)=5.45 is the appropriate
critical value from the studentized range distribution for pairwise
comparisons among 24 means based on an estimated standard error with
48 df for error.  Of course if you transformed the % adsorptions to 
stabilize variance, then you can ignore this remark.  

You could add the text in [brackets] to the captions of Tables 2,3:

b Means followed by same [lower case] letter within columns do not
differ significantly
c Means followed by same [upper case] letter within rows, for fixed 
oryzalin concentration, do not differ significantly.

I'm surprised that the adsorptions for the two pH values in Table 3 at 
resin=MN-400, conc=40000, do not differ significantly (95.4 and 81.3 
not different?)