[From Bruce Abbott (971210.1545 EST)]
I've got a lot to say in reply to a number of recent posts that in their
turn are replies to mine. Unfortunately, I don't have time to say it (my
little window of time-to-spare has closed with a bang). But I can't let
this one go by.
Bill Powers (971209.2023 MST) to me:
There has been only one person who has studied this article thoroughly
enough to realize what it shows (Isaac Kurtzer). Nobody else has made any
comments on it, other than "nice article." I started to show you how to use
this approach in analyzing your rat data, but you didn't seem very
interested.
I don't recall this at all. How did you propose to go about it? Why didn't
you try it yourself? Or did you? You have all the data . . .
Regards,
Bruce
[From Bill Powers (971210.1529 MST)]
Bruce Abbott (971210.1545 EST)--
I started to show you how to use
this approach in analyzing your rat data, but you didn't seem very
interested.
I don't recall this at all. How did you propose to go about it? Why didn't
you try it yourself? Or did you? You have all the data . . .
The first step was to estimate the unaccounted-for disturbances in the
environment function (unmeasured food spills, water balance, oxygen uptake
and CO2 expiration, etc.). The model we fit to the EFF was a leaky
integrator with loss rate proportional to weight and a conversion factor of
food intake into weight increments. If we assume this model, then we can
compute an equivalent food intake disturbance that would make up the
difference between the measured food intake and the intake needed to
produce the observed weight. I did try this, and found that the computed
disturbance was quite well-behaved for the animals that were not put
through the 17 or so different experimental conditions (at my last count)
during our explorations, and looked quite reasonable even for the
remainder. By that I mean that it changed slowly and amounted to only a few
grams per day. This is the part I was referring to when I said you didn't
seem very interested. No criticism intended -- we were having other
problems at that time, and I was also starting to lose interest in trying
to compensate for the difficulties in getting clean data.
The next step would have been to use this calculated disturbance in the
model, and then to calculate an equivalent reference signal from the data,
again assuming the correctness of the model. You were involved in your own
variations on the experiment by then, and I wasn't too enthusiastic about
going on because there were so many changes in experimental conditions that
the significance of the result would be questionable. Also I believe we
were having some serious doubts about whether the rats were actually
altering their rates of bar-pressing at all -- the initial concept of what
they were doing was sort of falling apart. We had abandoned relating
bar-pressing rate itself to intake and were looking at food delivery rate
as the "action" variable, which to me was a major setback. It didn't look
as though we really had found the aspect of behavior that was varying, and
if I recall properly, we also had questions about whether the measured
rates of pressing were being limited by the sluggishness of the lever
itself -- you were going to test a new lever.
I did learn a lot from this study. My main conclusion, barring some great
breakthrough, is that this sort of experiment is not a very good way to
find out what rats are controlling for, and how they do it. Perhaps
something could still be done with it, but the basic physical problems need
a lot of work.
Best,
Bill P.
It would be nice to try this method with a clean experimental run (no
changes from start to finish).
Best,
Bill P.