Please enable JavaScript.
Coggle requires JavaScript to display documents.
Quasi-Experimental Designs (none of these are true experiments) - Coggle…
Quasi-Experimental Designs
(none of these are true experiments)
Interrupted time series
Taking multiple measures before and after the event
Ex) M1, M2, M3, EVENT , M4 M5, M6
Ex) Measuring head injuries monthly 4 months before helmet law and monthly 4 months after helmet law
To control for this, you could introduce a nonequivalent control group (Basically like a control group of experimental design, but for a quasi design)
Ex) [context: helmet law] Find another similar place, With similar weather, That DID NOT change the law
However, even with nonequivalent control groups, there is still possibility for history effects like weather, or selection effects like not being able to randomly assign participants to conditions
Ex) non equivalent because we are not assigning participants to go to one high school or another; we don't know if there's something fundamentally different about these groups
Nonequivalent groups could help alleviate concerns for:
history (eg If a celebrity was in an accident)
testing (eg Both groups took two tests)
maturation & mortality (eg Both groups would be maturing and experience some drop out)
But they could not alleviate concerns for:
History (eg If a student from program school was in an accident)
Selection (eg We did not randomly assign students to groups)
Pretest-posttest
Only take SINGLE MEASURE BEFORE event and SINGLE MEASURE AFTER event
Ex) Take survey of attitudes about drunk driving beginning of junior year and end of junior year survey them again
Still cannot say causality since did not randomly assign students to comparison groups
Still has possibility for
History effects
Testing effects (being tested multiple times influences responses)
Exposure to the pretest may have caused students to rethink their drunk driving attitudes
Maturation (participants grow and develop, changed throughout the school year)
Mortality (Attrition) (participants drop out of study)
Developmental Studies
(age - a quasi - is variable studied)
Cross sectional
Between-subjects
Different participant groups are used at each age
Pro: Quicker and easier;
study can be conducted all at once
Con: Cohort effects
A cohort is a group of people with a shared experience or common identity not shared by others outside the group -- E.g., this quarter’s100B cohort, age cohort
Longitudinal
Within-subjects
Same participant groups are used at each age
E.g., Recruit 10-year-old participants and follow them for 60 years
Pro: Statistical strength Participants serve as their own controls
Cons: Long time to conduct study! Practical?
Mortality effects –participant dropout
History effects – outside events may influence results (outside events may influence results; it may not be of AGE; so we can NO say age is CAUSING)
Sequential
Combines aspects of Cross-Sectional and Longitudinal Designs
Fewer participant groups are recruited and followed across time
E.g., Recruit one set of participants who are10 and one set of participants who are 50, and follow them for 20 years
Quasi factorial experimental
Ex Post Facto
Prospective
Retrospective
Means AFTER THE FACT
find people who smoke and dont smoke; find out people to follow-- those who smoke 0, 20, and 30 a day; then identify how much develop lung cancer in future and see if there is a relationship, but canNOT say causal even if every smoker got lung cancer (since did not randomly assign)
Still can determine whether there is a relationship, but not that smoking CAUSES lung cancer
Con: Selection effects
.g., Could there be other differences between smoking and enon-smoking groups (Diet, education, doctor visits, etc.)?