Clarification on Ch8 -- also Solomon vs classic vs post-only control group designs
(Edited to add Hope's clarifications)
One-shot v. Static Group
- One Shot has ONE GROUP with stimulus and a post test
- Static Group Comparison has TWO GROUPS, one with the stimulus and both with post tests.
- They're both lousy. Neither has pretests.
The question came about because of Q9 on the homework about TV on emotional health in public vs Montessori school kids. Besides being lousy experimental design (what about emotional health as a result of going to different schools or one being public and one private and the economic status of families as a result of that?!!) I noticed that there's a lack of pretest so I'm assuming it's static-group comparison. However, it could also (I *think*) be a one-shot case study (why isn't it?). Because it's two groups.
---
I have some issues with 'natural experiments'.
In the example described on p233 about three-mi-island WHY is their 'natural experement' NOT classified as survey research. They say it's 'quasi-experimental and after-the-fact' and they use interviews and surveys to collect their data.
Is it because of subject selection differences? Help?!
---
Additionally, from the homework: Q15: Natural experiments are most likely to resemble which design?
a) static group comparison (YES?) (Correct, TWO GROUPS are compared, one with stimulus, one w/o, no pretests, both posttested).
b) classical (NO)
c) Solomon (NO)
d) 1group pre-post (No)
e) post-only control group (No, because of lack of randomization of subjects?)
---
Solomon 4-Group:
1: P, S, Po
2: P, _, Po
3. _, S, Po
4. _, _, Po
All have post-tests. Groups 1 and 2 have pretesting and are (exactly) classical design. Groups 3 and 4 lack pretests and are exactly Post-Test-Only Control group design.
Solomon allows comparison between control and experemental groups both with and without pretesting.
So: Classic + PoTCG --> Solomon
---
Why does classical theory suck? (Or, what's it bad at?). Can't control for interactions between pretest and stimulus (sensitization taking place).
Also, extra (unnecissary) expense of pretesting (and addl source of error) when can (theoretically) obtain same results with PostTest Only Control Group design because you've randomized assignment of subjects to control and experemental groups.
One-shot v. Static Group
- One Shot has ONE GROUP with stimulus and a post test
- Static Group Comparison has TWO GROUPS, one with the stimulus and both with post tests.
- They're both lousy. Neither has pretests.
The question came about because of Q9 on the homework about TV on emotional health in public vs Montessori school kids. Besides being lousy experimental design (what about emotional health as a result of going to different schools or one being public and one private and the economic status of families as a result of that?!!) I noticed that there's a lack of pretest so I'm assuming it's static-group comparison. However, it could also (I *think*) be a one-shot case study (why isn't it?). Because it's two groups.
---
I have some issues with 'natural experiments'.
In the example described on p233 about three-mi-island WHY is their 'natural experement' NOT classified as survey research. They say it's 'quasi-experimental and after-the-fact' and they use interviews and surveys to collect their data.
Is it because of subject selection differences? Help?!
---
Additionally, from the homework: Q15: Natural experiments are most likely to resemble which design?
a) static group comparison (YES?) (Correct, TWO GROUPS are compared, one with stimulus, one w/o, no pretests, both posttested).
b) classical (NO)
c) Solomon (NO)
d) 1group pre-post (No)
e) post-only control group (No, because of lack of randomization of subjects?)
---
Solomon 4-Group:
1: P, S, Po
2: P, _, Po
3. _, S, Po
4. _, _, Po
All have post-tests. Groups 1 and 2 have pretesting and are (exactly) classical design. Groups 3 and 4 lack pretests and are exactly Post-Test-Only Control group design.
Solomon allows comparison between control and experemental groups both with and without pretesting.
So: Classic + PoTCG --> Solomon
---
Why does classical theory suck? (Or, what's it bad at?). Can't control for interactions between pretest and stimulus (sensitization taking place).
Also, extra (unnecissary) expense of pretesting (and addl source of error) when can (theoretically) obtain same results with PostTest Only Control Group design because you've randomized assignment of subjects to control and experemental groups.
posted by Clstal at 5:18 PM
3 Comments:
One-shot v. Static Group
- One Shot has only ONE GROUP used with an experiment and a posttest
- Static Group Comparison has TWO GROUPS with one receiving the Independedn Variable and both groups being posttested and compared.
You're right - they're both lousy. Neither of these have pretests. The correct answer then is a Static Group Comparison because we are using TWO GROUPS and comparing them.
The Three Mile Island was so flawed it's amazing they tried to draw any conclusions in the first place. I don't think I understand your question though. The survey of the people was the foundation of the "natural" experiment raises many validity problems as discussed on page 244, but I don't understand the question of "WHY is their 'natural experiment' NOT classified as survey research."
Q15: Natural experiments are most likely to resemble which design?
a. static-group comparison is correct since TWO GROUPS are compared, one having been subjected to the independent variable and the other not. Neither are pretested, and both are posttested.
THANK YOU for the clarifications (I added em to the original).
The natural experement thing I think is just my not understanding why it's even classified as an experiment...? To me, it should be a survey, both self-administered and interviewer-administered.
To me experements are all about control and careful setup and subject randomization, etc. With how strong they are on internal validity and how weak on external validity and by def have introduced stimuli I don't see how 'disaster research' even qualifies as 'experemental' since it's after-the-fact and has minimal control and no subject randomization. It dosn't have even similar issues of validity and seems like it should be a special type of survey research rather than a special type of experemental research.
(Or at the very least, they should have put it under 'quasi-design' as an example of why quasi-designs are sometimes necissary.)
My only concern is that I'm somehow missing some aspect of 'natural research' that makes it fit the experement model much better.
Your understanding seems to be about the same as mine on the "natural" experiments. I agree that the wording screws the whole concept up, but the way in which the research/experiment is conducted falls right in with the Pre-Experimental Designs (which, once again, suck)and I think that's how they justify referring to them as "experiments".
Post a Comment
<< Home