Chapter 12: Experiments and Observational Studies Flashcards
Confounding Variable
Another variable that is related to the explanatory variable and influences the response variables and may create a false perception of association between the two
Experiment
Treatments are intentionally imposed on experimental units
Explanatory Variables (Factor)
May help predict a change in the response variable
Response Variable
Used to measure the outcome of a study
Components Of A Well-Designed Experiment
- Comparison (Blocking) of at least 2 treatment groups, one of which could be a control group
- Random assignment/allotment of treatments to experimental units
- Replication (use enough experimental units in each treatment group)
- Control of potential confounding variables, where appropriate
Random Allocation
- Number experimental units 1 through N
- Select random numbers, put the first unit in group one, second in group two, and cycle through again once desired amount of groups has been reached
- Continue until desired number of experimental units in each group is reached
Randomized Block Design
Ensures that units within each block are similar with regard to a blocking variable; Not randomly done; Randomization is performed within each block after blocking is done
Blocking Variable
Characteristic of experimental units used to create groups, or blocks, of similar units; Helps separate natural variability from differences due to the blocking variable
Placebo
Fake treatment that is similar to the treatments being tested
Placebo Effect
Occurs when experimental units have a response to a placebo
Blinding
Occurs when the subject and/or the researchers are unaware of the treatment being administered
Single-Blind Experiment
Subjects do not know which treatment they are recieving, but researchers do (or vice versa)
Double-Blind
Neither the subject nor the researchers who interact with subjects are aware of the treatments being administered
Matched Pairs Design
- Special type of block design (block size 2)
- Arranged such that the blocks are closely aligned
- Within each block, both treatments are randomly assigned
- (Alternate: each subject may receive both treatments)
Statistical Significance
Statistically significant differences between or among treatment groups are evidence that the treatments caused the effect (if random assignment)
Statistical Inference
Allows us to make decisions about the distribution (population) of treatments of interest based on the results from the sample; If experimental units are representative of the population, then the results can be generalized to the population of subjects like the ones in the study; Random selection of individuals gives a better chance that the sample will be representative of the population
Association/Relationship
General term meaning there appears to be some relationship between variables. This can be used with any combination of categorical and quantitative variables; Appropriate for observational studies and experiments
Correlation
Precise term describing the strength and direction of a linear relationship (correlation coefficient, r). Can only be used with 2 quantitative variables; Appropriate for observational studies and experiments)
Causation
Changing the explanatory variable is not only “associated” with changes in the response variable, it causes those changes; Appropriate for experiments only
Scope Of Inference
Refers to the type of inferences (conclusions) that can be drawn from a study. The types of inferences we can make (inferences abut the population and inferences about cause-and-effect) are determined by two factors in the design of the study:
- Random sample & random assignment: Generalize to population & determine cause and effect
- Random sample & not random assignment: Generalize to population, but not determine cause and effect
- Not random sample & random assignment: Not generalize to population, but determine cause and effect
- Not random sample & not random assignment: Not generalize to population & not determine cause and effect
Observational Study
Survey that does not impose treatments (cannot determine cause and effect from this type of study)
Retrospective
Examines current/past data of individuals
Prospective
Examines data that follows the individuals into the future
Completely Randomized Design
Experimental units are assigned to treatment groups completely at random
