At 05:06 PM 3/1/00 -0500, you wrote:
>
>how necessary is it to randomly select animals for the control and
>experimental groups in animal studies?
>so you got 100 rats,50 to go to the E group and 50 to go to the
>C group,can randomization really equalize for the two conditions?
>It appears that there is already a stability factor,since some of those
>animals may be bred from the same strain or litter?
>
>Send me something.
>
>Michael Sylvester
>Daytona Beach,Florida
While similar in genetic background, in environmental background, in age
and size, there still are individual differences among animals. (Also note
that even if they were all cloned siblings, we still could not control
their environmental experiences so well that we could ensure that they were
identical. And as I understand it even clones show some individual
differences.) So, we have two choices to equate the treatment conditions.
(1) Used a within-subjects design, counterbalancing for order of the
treatments, which would ensure that the individual differences were
controlled by holding them constant across treatments. (2) Use a
between-subjects design, with the subjects randomly assigned to the
treatments. Does random assignment guarantee that the treatment conditions
are identical? No, of course not. Chance factors always could result in
bias in the groups. But especially with the random assignment of a sample
size of 100 (50/condition), the likelihood of there being any systematic
variation between the groups with respect to any particular trait of the
subjects is dramatically reduced. Also note that as the sample size is
increased, the liklihood of systematic variation between the groups is
reduced. (And it seems to me that beyond about 30 subjects per treatment
condition, further increases in sample size yield only negligible benefits
with respect to eliminating chances of systematic variation. Anyone have
any specific info on that?)
Michael, if you don't do either (1) or (2), you have a confound.
This sounds like a good question for my next research methods exam. (And
given how much I hammer away at just this point I would hope and expect my
students to be able to answer it!)
Bob
****************************************************************
Robert T. Herdegen III
Department of Psychology
Hampden-Sydney College
Hampden-Sydney, VA 23943
****************************************************************
