How to Design and Conduct an Experiment CONTROL

  • Slides: 7
Download presentation
How to Design and Conduct an Experiment • CONTROL: control lurking variables by comparing

How to Design and Conduct an Experiment • CONTROL: control lurking variables by comparing several treatments in the same environment • RANDOMIZATION: random assignment of subjects to treatment groups, helps even out effects of lurking variables which helps avoid bias • REPLICATION: using a large enough number of subjects, or repeating the experiment, to reduce chance variation

Vocabulary and Application • TREATMENTS: imposed condition on subjects to test for differences in

Vocabulary and Application • TREATMENTS: imposed condition on subjects to test for differences in response • CONTROL GROUP: a group with which you compare effects of treatment • EXPERIMENTAL UNITS: individuals being tested, called subjects if human • RANDOM ASSIGNMENT: ways to randomly assign treatments to negate bias

• BIAS: certain outcomes are systematically favored and results are overestimated or underestimated

• BIAS: certain outcomes are systematically favored and results are overestimated or underestimated (ie: undercoverage, nonresponse, response bias-lying, wording of question, characteristics of interviewer, anonymity) • CONFOUNDING: when it is impossible to separate the effects of two variables • LURKING: effecting the response but not measured in the experiment • PLACEBO EFFECT: when you think you should be effected and so you act affected • BLINDING: double or single, when the subject, tester, or both doesn’t know which treatment is being imposed

Types of Experimental Design COMPLETELY RANDOMIZED DESIGN: when all experimental designs are randomly assigned

Types of Experimental Design COMPLETELY RANDOMIZED DESIGN: when all experimental designs are randomly assigned to a treatment (ie: all drawn out of a big hat) Treatment Placebo Vaccine 500

BLOCK DESIGN: 1) divide subjects into blocks based on characteristics that could effect outcomes

BLOCK DESIGN: 1) divide subjects into blocks based on characteristics that could effect outcomes 2) randomly assign treatments within each group 3) compare results within each block Treatment Placebo Vaccine Male 250 Female 250 Gender

MATCHED PAIRS DESIGN: pair two similar people together and randomly assign one treatment to

MATCHED PAIRS DESIGN: pair two similar people together and randomly assign one treatment to each, or if possible give each person both treatments in a random order Treatment Placebo Vaccine 1 1 1 2 1 1 . . 499 1 1 500 1 1 Pair

SOURCES • NOTES • AP STATS TEST TAKING TIPS, SUMMARIES, FORMULAS, AND MORE •

SOURCES • NOTES • AP STATS TEST TAKING TIPS, SUMMARIES, FORMULAS, AND MORE • STATTREK. COM