Part 3 - Quasi Experimental Design Flashcards
What is a quasi-experimental design?
Experimenter does not design experiment but takes naturally occuring processes (i.e. time series, different natural groups) - lacks full control over scheduling of experimental stimuli
Discuss a time-series experiment?
many observations to account for history w/o control group
BUT: Problem of internal validity - could anything else have explained the observations?
What is a Equivalent Time-Samples Design and what might be problems?
- Design is a form of the time-series experiment with repeated introduction of the experimental variable
- History is controlled by presenting X on numerous separate occasions; any rival explanation based on coincidences with extraneous events becomes unlikely
- Generalization only possible to frequently tested populations, or: problem of “multiple-X interference”
What is a Nonequivalent Control Group Design and what might be problems?
O O
- used in many instances in which true experimental designs = randomization not possible)
- must hypothesize an interaction between history, maturation, or testing and the specific selection differences that distinguish the experimental and control groups
- Often unlikely, there are a number of situations in which interactions might be invoked
What are Counterbalanced Designs (or “rotation experiments”) and what might be problems?
- All respondents are confronted with all treatments
- Treatments are applied in a randomized manner
- Design contains three classifications: groups, occasions, and Xs; each classification is “orthogonal”: each variate of each classification occurs equally often
- Problem: differences between effects of Xs could be instead a specific complex interaction effect between group differences and occasions
What is a Separate-Sample Pretest-Posttest Design and what might be problems?
R O (X)
R X O
•Large populations, where random segregation of subgroups is not possible, but one can exercise something like full experimental control over the whenand to whom of the O, employing random assignment procedures
•One sample is measured prior to the X, an equivalent subsequent to the X
•Main weakness: failure to control history
•In principle very high external validity
What is the idea behind experimental economics?
well planned incentives in the experiment induce decisions where personal characteristics of respondents as well as “error terms” become less relevant
What are the three conditions according to Smith to induce decisions and to secure experimental control?
monotonicity, salience, dominance
What is monotonicity?
Individuals must prefer more payment to less
-> Problem: empirical findings from dictatorship game do not support this
What is salience?
The payment must depend on the individuals’ decisions, and the respondents must be enabled to understand the underlying relationship
What is dominance?
changes in the utility level of the respondent must (primarily) result from the experimental payments and not from any other factors that are beyond experimental control
-> Problem: framing
Why would you include decision preparation phase?
addressing bounded rationality:
– advantage of more rationality in behavior
– disadvantage of less “natural”, realistic decisions
– thus highly debated
What are clinical decision situations and why might they be criticized?
clinical” decision situations characterized only by payments and without any reference to real decision situations
- tradeoff between external validity and meeting the dominance criterion (that may be judged as important for internal validity since it rules out alternative explanations)?
