1. Statistics 200 Lecture #10 Thursday, September 22, 2016
Textbook: Section 5.2, 6.1, 6.2, 6.3
Objectives:
• Use margin of error formula to find sample size needed for particular MOE
• Understand and identify various types (prospective, retrospective, case- control) of observational studies
• Understand and identify various techniques (blinding, placebo, pairing/blocking) used in randomized experiments
• Understand ethical reasons for preferring observational studies over randomized experiments in some cases
• Identify explanatory, response, and confounding variables in context.

2. Skills from this lecture
Identify!
Explanatory
Response
Confounding
variable types
Identify presence and type of blinding, placebos.
Matched pair and block designs
Randomized Experiment
Retrospective vs. Prospective
Identify cases and controls in
case-control studies
Observational Study

3. Last lecture we discussed…
Surveys
Difference between population and sample
Difference between survey and census
Three types of bias
Confidence Intervals
Calculate p-hat, margin of error
Construct a conservative 95% confidence interval for a population proportion
Find sample size for desired margin of error

4. What proportion of dog owners have more than one dog?
The confidence interval is 14% plus or minus 5%.
We are 95% confident that the true population proportion of dog owners with more than one dog is between ____ and _____.
.09
.19
Clicker Question 1: Is it reasonable to conclude that fewer than 15% of dog owners have more than one dog?
Yes No

5. What proportion of dog owners have more than one dog?
The confidence interval is 14% plus or minus 5%.
We are 95% confident that the true population proportion of dog owners with more than one dog is between ____ and _____.
.09
.19
Clicker Question 2: Is it reasonable to conclude that fewer than 20% of dog owners have more than one dog?
Yes No

6. Using desired margin of error to choose n
We can design our survey to have a margin of error of our choosing by specifying the sample size, n.
Example: if desired ME is 0.02,
= 0.02
Solve for n

7. Today’s Objectives: Revisit in detail…
difference between observational study and randomized experiment
Identification of explanatory and response variables
Introduce…
treatments and experimental units
Ideas behind experimental design
Ideas behind design of good observational studies.
Describe the following types of experimental designs: Completely Randomized, Matched-pair and Randomized Block

8. You may remember…
We talked about how you cannot make a cause-and-effect conclusion based on an _________________, but you can make one on the basis of a __________________
We’re going to delve further into these types of studies.
observational study
randomized experiment

9. Does exercise increase exam scores?
Motivating example:
Does exercise increase exam scores?
Two ways to conduct a study to answer this question
Observational
Experimental
Randomly select folks to participate
Record responses about exercise frequency and exam scores during fall semester
Randomly select folks to participate
Randomly assign exercise levels for this semester and then record exam scores

10. Observational Studies
In an observational study, the researchers simply _______ or ________ the participants about opinions, behaviors, and outcomes.
Researchers do not ____________ any treatments or conditions
Participants of observational studies aren’t necessarily people, they can be any type of observational unit.
question
observe
assign

11. Consider our example: Does exercise increase exam scores?
Randomly select folks to participate
Record responses about exercise frequency and exam scores during fall semester.
Random selection of participants – good.
No assignment of anything to these participants—hence observational.

12. Strengths and weaknesses of observational studies
Usually, observational studies are….
cheaper
faster
Weaknesses
No cause and effect because there may be
variables behind associations.
confounding

13. Confounding variable
A variable that is related to the explanatory variable and also affects the response variable is called a confounding variable.
aka a ______ variable
1lurking
The effect of a confounding variable on the response cannot be separated from the effect of the explanatory variable.

14. Confounding variable example:
After collecting a large amount of data, a researcher found a very interesting relationship between ice cream sales and the number of drownings in towns on the East Coast.
When ice cream sales increase, so do the number of people drowning.
Confounding variable:
Temperature

15. Types of observational studies
Retrospective
Uses data from the _________.
past
Example: motivating study on exercise and exam scores
Prospective
Follow participants into the _________.
future
Example: The Up 7 documentary series.

16. Case-control studies
Cases, who have a particular attribute or condition, are compared to Controls who do not.
Often used in medical settings. Good case-control observational studies try to identify controls who are as similar as possible to the cases except that they don’t have the disease
Example: Compare genetic makeup of family members that have Parkinson’s disease to that of family members who do not have Parkinson’s disease.
Cases

17. Case-control studies
Cases, who have a particular attribute or condition, are compared to Controls who do not.
Often used in medical settings. Good case-control observational studies try to identify controls who are as similar as possible to the cases except that they don’t have the disease
Example: Compare genetic makeup of family members that have Parkinson’s disease to that of family members who do not have Parkinson’s disease.
Cases
Controls

18. Cause-and-effect conclusions are usually desirable
We cannot claim cause-and-effect from an observational study.
For this we need a…
randomized experiment!

19. Randomized Experiments
If we really want to show a causal relationship, we need to use a randomized experiment.
In a randomized experiment, participants are randomly assigned to participate in one condition or another.
These different conditions are called __________.
treatments

20. Consider our example: Does exercise increase exam scores?
Recruit study participants
Randomly assign exercise levels for this semester and then record exam scores
Need to get volunteers and informed consent
Random assignment of treatment (exercise level). The randomness is key.

21. Strengths and weaknesses of randomized experiments
Can make
cause-and-effect conclusions
Weaknesses
Usually take more time
Usually require more $$$.
Sometimes, treatments cannot be randomly assigned for ethical reasons.

22. Ethical reasons informed consent
The big reason that randomized experiments aren’t always used is because they can put people’s welfare at risk.
Sometimes, when there is some risk involved in the treatment, subjects will be required to sign __________________ forms, in which they are introduced to some of the risks associated with the experiment.
informed consent

23. Randomization: the crucial element
We do experiments to reduce the chance that our results will be affected by confounding variables and bias. Randomization is key.
Randomizing type of treatment
Randomizing order of treatment
If each experimental unit receives all treatments, we can randomize the order
Which treatment each experimental unit receives.

24. Control groups placebo
When we aim to show that a certain treatment has an effect, we must know what would happen without it.
Typically, a control group is used. This group is treated exactly the same as the other group(s), except it is not given any active treatment.
Often, control groups are given a _______, which seems like a treatment, but is actually inert.
placebo

25. Control groups and blinding
_____________ experiments are experiments in which the participants don’t know whether they are receiving the placebo or the treatment.
_____________ experiments are experiments in which neither the participants or the researchers know whether a participant is receiving the placebo or the treatment.
Single-blind
Double-blind

26. Repeated-measures design
Pairing and blocking
• Sometimes, giving a subject both the active treatment and the control can help to better determine the treatment effects.
Repeated-measures design
• Sometimes, we match people together based on a certain characteristic, like BMI or IQ and give one of them a control and one of them the active treatment.
Pairing is a special case of blocking.
Matched-pair design

27. Pairing and blocking Young All Participants Middle-aged Elderly
Block design:
Divide participants into similar groups, then randomly assign treatment / control within each group
Random treatment assignment
Young
All Participants
Random treatment assignment
Middle-aged
Elderly
Random treatment assignment

28. Pairing and blocking - conclusion
When we talk about experiments:
If the experiment has randomly assigned treatments without pairing or blocking, that experiment would be said to have a ________________________________.
If the experiment uses pairing, it is said to have a _______________________.
If the experiment uses blocking, it is said to have a __________________________.
completely randomized design
matched-pairs design
randomized block design

29. Example 1 Explanatory: Caffeine (yes/no) Response: Swimming time
A researcher wants to know whether caffeine will increase swimming speed. She randomly assigns 25 volunteers to take a caffeine pill, and the remaining 25 to take a placebo, then compare the swimming times of the two groups.
Explanatory: Caffeine (yes/no)
Response: Swimming time

30. Example 2 Explanatory: Caffeine (yes/no) Response: Swimming time
Each of 50 participants randomly takes one treatment, measures their swimming speed, then returns the next week to take the other treatment and measure swimming speed. The measurements for each participant are compared.
Explanatory: Caffeine (yes/no)
Response: Swimming time
What type of design is this?
Repeated measures
Blocked
Paired
Completely randomized

31. Review: If you understood today’s lecture, you should be able to solve
6.1, 6.3, 6.7, 6.11, 6.13, 6.17, 6.19a-d, 6.27, 6.29, 6.31, 6.35, 6.45, 6.89
Objectives:
• Use margin of error formula to find sample size needed for particular MOE
• Understand and identify various types (prospective, retrospective, case- control) of observational studies
• Understand and identify various techniques (blinding, placebo, pairing/blocking) used in randomized experiments
• Understand ethical reasons for preferring observational studies over randomized experiments in some cases
• Identify explanatory, response, and confounding variables in context.