sir_toejam
Posts: 846 Joined: April 2005
|
Quote (Wesley R. Elsberry @ Mar. 30 2006,09:38) | The item in the methods section that I saw as a problem was that the group of treatment fish was so small. I see that in the discussion, you note that specific thing as a reason to treat all results as preliminary. But I didn't see a way to state that in one word, so I think you must be referring to something else. |
ack! sorry for being so long answering the response; I was gone for a while and hadn't seen any responses for at least a couple of weeks prior, so i figured there hadn't been any interest and forgot all about it.
you're red hot, but you must have covered the answer with your foot
not just small, but singular!
only one treatment fish of each type was used.
the term I'm sure you recognize for this is:
"pseudoreplication".
for those unfamiliar with the term, it basically refers to an experimental design where there are multiple variables, but one or more of them is insufficiently replicated within the design of the experiment.
for a more in depth discussion of what this is and the prevalence of it in the literature, here's an abstract of a nice little review paper (which i was forced to read by my advisor after i published that paper ):
http://www.esajournals.org/esaonli....ge=0187
In this specific case, there are two primary variables we are looking at, the behavior of the adults, and the color of the treatment fish.
to control for randomness (and help control for other potentially competing variables), we replicate, of course, and the adult subjects were replicated (sufficient at least for a non-parametric analysis - ask if you want me to elucidate the difference between non-parametric and parametric), but the treatment fish were not. This introduces the possibility that the specific treatment fish may have been abberations.
Ideally, in a better design we would replicate the treatment fish as well, in order to control for this.
The only justification for non-replication of treatment fish in this experiment was that prior to this study, most all similar studies of this type were conducted with painted model fish (again, no replication). I figured at the time that the general color of the animal was the most likely trigger for variable behavior in the adults (that's what I was testing, after all), but of course in hindsight, there were other possibilities....
any guesses?
and, thanks for expressing interest. always nice to know someone read it.
cheers
|