Naive question:
I have two populations treated and untreated, with several
measurements taken before and after treatment, e.g. of the format
grams of food eaten on day one of month
Month -2 -1 0 1 .... n
Rat 1 10.2 10.5 4.1 5.2 ...
Rat 2 8.1 7.2 6.5 7.0 ...
Rat 3 7.2 8.3 6.9 6.8 ...
I can do T or U tests at each timepoint comparing treated and
untreated to check for a significant effect, but would like an
aggregate measure (something like Area Under Curve) for post-treatment
measurements that does not ignore the sequential nature of the
measurements or the variability. The timeframe of the measurement is
much less than the time interval between each test (i.e. grams eaten
one day out of a month, rather than food eaten per month, which would
be a better measurement).
I was naively thinking of using the area under curve after time 0
(start treatment) and treating the AUC as a single data point, then
doing a T or U test between the treatment/control groups of AUCs,
after testing normality. This sounds like it will discard the
sequential nature of the data. The treatment may result in a
temporary dip in grams eaten but is not expected to be linear with
time (there will probably be a compensation at some later time).
There is likely a better test I am unaware of (as you can tell, I'm
not a statistician ;) ). Any ideas?
Thanks for any help,
Daniel S.
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
