Expts: tell me about em and use as much termanology as possible
So Experiments are the thing, so far, that I'm most rusty on... so tell me about em...
What part of the design process are there? What decisions are to be made? Why would you make those decisions? What're they useful for and why? What's some expt-related terminology that you think is important?
Would you ever want to run one? Why? What'd you want to test? How would you do it? Why?
--Crystal
What part of the design process are there? What decisions are to be made? Why would you make those decisions? What're they useful for and why? What's some expt-related terminology that you think is important?
Would you ever want to run one? Why? What'd you want to test? How would you do it? Why?
--Crystal
posted by Clstal at 11:03 AM
1 Comments:
So. To partially respond to my own post:
I'm interested in sleep and it's effect on performance. I'd want to run a study comparing my vol's sleep patterns and their performance. Strictly exploratory (not introducing any stimulus besides the exptal conditions, which would be significant). I want to know what normal sleep looks like. I also want to know the big fat why, but don't expect to learn that except through years of research (which I won't be doing) -- so mostly I'll be following along watching the field eagerly.
I'd use a multipart format with surveys for my control group (detailed, lengthy, paid, mail surveys in their own homes after an orientation session) attempting to elicit the same info from them as I'm getting directly (through sleep monitors and performance measuring tests as well as surveys) from my expt group.
Subjects would be put in exptl or control group through a randomized matched process similar to the one on p224 for an educational enrichment expt.
The results generalizability would depend on my expt size and in the quality of information that I'd be able to gather, and also, significantly, on the sophistication of the software for the performance testing. (IE testing response speed is of only minimal use if you're trying to say that business execs would perform better with more sleep, more accurate would be software that tested their ability to think and formulate complex responses and handle stress and multitask and feel confident).
This was a lousy example and dosn't fit one of Babbie's expt models. :-( The closest would be the 'posttest-only control-group design'. The stimulus would actually be the exptl measuring 'stuff' (perf-test software and sleep-measuring wristband) and both groups would get the same surveys.
Also, the goal wouldn't be to compare the two groups for differences but hope that both groups provided the same data (and therefore that the results of the exptl group were generalizable to a larger pop. IE testing the H0 as the H1 (from stats with Coventry) and HOPING to prove the 'no difference' hypothesis true.
Helpful or bewildering navel-gazing?
Post a Comment
<< Home