Statistical Sciences 2037A/B Chapter Notes - Chapter 5.1-5.5: Dependent And Independent Variables, Confounding, Random Assignment
![](https://new-preview-html.oneclass.com/rKZ3w4gpqWELjZVEkyLqj8nX2G5YyBMv/bg1.png)
Monday, October 17, 2016
Study Design
5.3 Difficulties and Disasters in Experiments
-Some potential complications:
1. Confounding variables
•Problem: variables that are connected with explanatory variable can distort the
results of an experiment; may be the agent actually causing a change in the
response variable, and not the explanatory
•Solution: randomization: if experimental units are randomly assigned to
treatments, effects of confounding variables should apply equally to each
treatment, thus observed differences should not be attributable to confounding
variables
2. Interacting variables
•Problem: second variable can interact with the explanatory variable, but results
are reported w/o taking that interaction into account = reader mislead into
thinking treatment works equally well, not matter what the condition is for the
second variable
•Solution: researchers measure and report variables that may interact with main
explanatory
3. Placebo, Hawthorne, and experimenter effects
•Problem:
-placebo: power of suggestion is somehow able to affect the result
-hawthorne effect: participants in an experiment response differently than
they would otherwise, just bc they are in the experiment
-experimenter: numerous ways they can bias the results
•Solution: can be overcome by using double-blind designs and by including
placebo group or a control that receives identical handling except for the active
part of treatment
!1
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
Placebo: power of suggestion is somehow able to affect the result. Hawthorne effect: participants in an experiment response differently than they would otherwise, just bc they are in the experiment. Experimenter: numerous ways they can bias the results: solution: can be overcome by using double-blind designs and by including placebo group or a control that receives identical handling except for the active part of treatment. Researchers can only observe and not control explanatory variables. Advantage that they are more likely to measure participants in their natural setting. Types of observational studies: case-control studies. Cases" who have a particular attribute or condition are compared with the. Cases and controls are compared to see how they differ on the variable of interest. Sometimes cases are matched with controls on an individual basis: retrospective or prospective studies. Retro: participants are asked to recall past events.