Principles of exp. design

1
Principles of exp. design

• Control for effects of lurking variables
• Randomization to keep personal biases or other
  preferences out of the study
• Replication of the experiment reduces chance
  variation

2
Completely randomized design

• All experimental units are allocated at random
  among all treatments.
• Ex Clinical trial to test effectiveness of new
  drug vs. current standard for alleviating
  migraine headaches. Test subjects are people
  known to have migraine headaches.

3
More complicated designs

• In some cases, we can use the idea of
  randomization to develop more complex and more
  efficient designs
• There are many special cases well touch on
  matched pairs and block designs

4
Matched pairs designs

• We can sometimes match an experimental unit of
  the control group with one from the treatment
  group (can extend to more groups)
• Ex Studies with identical twins as volunteers
  can randomly assign one to each treatment (or
  control)

5
Block designs

• Similar experimental units assigned to same block
• Blocks differ in ways which may affect experiment
• Ex In agricultural field trials, each plot of
  similar soil type may form block
• Random assignment of treatments made within each
  block

6
Probability/inference
Use inference
Population
Sample
Use probability
7
Population vs. sample

• Population is entire group we want to get
  information about
• Sample is subset of population that we are able
  to survey or examine
• Make inference about population by looking only
  at sample
• Sample design is method used to choose sample

8
Voluntary response sample

• Data/responses from people who respond to a
  general appeal/request
• Tends to be biased, because respondents are
  mostly those who have time to respond or are
  motivated by strong opinions
• Not necessarily indicative of whole population

9
Survey problems

• Undercoverage some groups in the population are
  left out of the process of choosing the sample.
• Nonresponse individual chosen for the sample
  cant be contacted or does not cooperate.
• Response bias behavior of interviewer or
  respondent can influence responses.
• Wording of questions
• Desire of respondent to look good

10
Simple random sample

• Is special case of probability design, which
  gives each individual some chance of being chosen
• Sample size usually denoted by n
• Sample chosen from population so that each subset
  of size n has same chance of being selected
• Since each individual has same chance of being
  chosen, theres no systematic bias

11
Stratified random sample

• By chance, a simple random sample may not get
  enough people from important groups.
• Solution take a simple random sample of each
  group (stratum) separately
• Combine simple random samples to obtain a
  stratified random sample (similar to blocking)

12
Multistage samples

• Hard to get a simple random sample on a large
  scale (how to choose, how to physically get there
  for interview, etc.)
• Solution choose sample in stages
• Randomly choose areas (for instance, counties),
  then further subdivide (townships)
• Subdivide into neighborhoods or other small areas
• Take sample of houses in these areas

13
Survey problems

• Undercoverage some groups in the population are
  left out of the process of choosing the sample.
• Nonresponse individual chosen for the sample
  cant be contacted or does not cooperate.
• Response bias behavior of interviewer or
  respondent can influence responses.
• Desire of respondent to look good
• Wording of questions

14
Dealing with non-response

• Persistence
• If at first you dont succeed
• Methodology
• Which methods below do you think will have the
  biggest problem with non-response?
• Telephone interview
• Mailings
• Email survey
• Face-to-face interview

15

• Some cell phone users have developed brain
  cancer. Should all cell phones come with a
  warning label explaining the danger of using cell
  phones?

16

• Do you agree that a national system of health
  insurance should be favored because it would
  provide health insurance for everyone and would
  reduce administrative costs?

17

• In view of escalating environmental degradation
  and incipient resource depletion, would you favor
  economic incentives for recycling of
  resource-intensive consumer goods?

18

• Which of the following best represents your
  opinion on gun control?
• The government should take away our guns.
• We have the right to keep and bear arms.