Chapter 6 Experiments in the Real World PDF

Preview

Summary

Stewart Hamilton...

Description

Chapter 6: Experiments in the Real World Equal Treatment for All ● Randomized comparative experiment ○ All the subjects are treated alike ○ The treatment’s effect is being evaluated (compared) not the differences in the participants ● An Experimenter Must ○ Know exactly what information is wanted and must compose questions to extract the information from the sample ○ Know exactly the treatments and responses (categorization) ○ Sometimes Construct the apparatus needed to apply the treatments and measure the responses ● Single-blind experiment: subjects do not know which treatment they are receiving ● Double-blind experiment: neither subjects nor evaluators know which treatment was given to which subjects ○ Used extensively in medical experiments ■ Example: Aspirin Study ● Placebos Work! ○ Medical studies must take special care to show that a new treatment is not just a placebo ● More randomization ○ Example: Rats ● Refusal: Same Issues as with Sampling ○ Bias can result if those who refuse are systematically different from those who cooperate ■ E.g., Minorities, women, the poor, and the elderly are often underrepresented in Medical Experiments and have low rates of participation in clinical trials ● Non-adherers: Subjects who participate but do not follow the experimental treatment ● Dropouts: Experiments over an extended period of time will likely have subjects who begin the experiment but do not complete it ● Well-designed experiment ○ Changes in the explanatory variable ○ Cause changes in the response variable Were the Treatments and Responses Realistic? ● Treatment ○ Experiment may involve artificial situation ■ E.g., sleep room ○ Tested on animals ■ E.g., Foods that cause cancer are tested in large doses over a short period of time

○ Responses ■ Subjects may not respond as they would in real life ■ Frustration level of individual over time ○ Hawthorne Effect: Also referred to as the “observer effect” ■ Individuals improve (or change) an aspect of their behavior in response to their awareness of being observed ■ Behave different when watched Experimental Design in the Real World ● Completely Randomized Design ○ The experimental subjects are allocated at random among all the treatments ● Questions ○ Did the subjects volunteer to participate? ○ Are the subjects representative of a larger population? ○ Do the subjects differ from the general population? ● A completely randomized design can have any number of explanatory variables ● Goal: Obtain more precise results by using more sophisticated designs in experiments ● Three Basic Designs: ○ Completely Randomized Design (Section 1.2) ■ All the experimental units are allocated at random among all the treatments ○ Block Designs (New) ○ Matched-Pairs Designs (New) Basic Ideas of All Three Designs ● Control the effects of lurking variables on the response ○ Most simply by comparing several treatments ○ Gain additional control by using block or matched-pairs designs ● Randomization: random allocation of units to treatments with the purpose to: ○ Eliminate bias due to systematic differences in the groups ○ Use laws of probability to draw conclusions from data ● Replication of the experiment on many subjects ○ Purpose: reduce chance variation in the results Block Design ● Blocking is similar to stratification ○ Stratification: know population and see if sample is the same or not ■ How to pick ○ Block: make sure subject are split within sample ■ How to treat ● Block: ○ Group of experimental subjects that are known before the experiment ○ Similar in some way that is expected to affect the response to the treatments ● Random assignment of subjects to treatments is carried out separately within each block

● Randomized block experiments are more precise than completely randomized experiments Matched Pairs Design ● Matched pairs are block designs ○ Compares just two treatments ○ Does not randomly assign subjects to the two treatments prior to matching ● Choose pairs of subjects as closely matched as possible ○ This allows matching to reduce the effect of variation among the subjects ● Randomize only within each matched pair ○ Randomly assign treatments to each subject in a pair ● Example: Are right hands stronger than left ○ Completely Randomized Design ○ Squeeze scale between thumb and finger ○ Measure results and draw conclusion ■ Issues: • Right Handed vs. Left Handed ○ Better Questions ○ Is the right hand of right-handed people generally stronger than the left? ○ Is the dominant hand generally stronger than the non-dominant? ○ Is the right hand of right-handed people generally stronger than the left hand of left-handed people?...