Interleaves

Question 1

What is pseudoreplication?

Answer 1

The error that occurs when samples are not independent, but they are treated as though they are.

Question 2

How can pseudoreplication be avoided?

Answer 2

Avoid by summarizing data on individuals and remembering that sampling units may be larger than the individual.

Question 3

What are the four steps to making a plan for confirmatory research?

Answer 3

1. Develop clear statement of the research question
2. List the possible outcomes
3. Develop experimental plan - keep it simple
4. Discuss the design with others

Question 4

What is the breakdown of step 1: develop clear statement of the research question?

Answer 4

• Why interesting/important?
• What’s already known?
• What contrasting predictions do the hypotheses make?

Question 5

What is the breakdown of step 2: list the possible outcomes?

Answer 5

• Can you draw firm conclusion no matter what the outcome?
• Simulate data to explore possible outcomes
• What is the desired scope of inference?

Question 6

What is the breakdown of step 3: develop experimental plan?

Answer 6

• How many treatments?
• What effect size is meaningful?
• Necessary sample size for desired power?
• Confounding variables?
• Avoid pseudoreplication

Question 7

What is the breakdown of step 4: discuss design with others?

Answer 7

• Without proper consultation, studies can be dead on arrival.

Question 8

Why are controls important?

Answer 8

A drug had an 88% success rate and a placebo group was also tested with 40% showing similar improvement. Drug was effective but not to the same register as originally thought.

Question 9

What is the placebo effect?

Answer 9

The improvement in medical condition that results from the psychological effects of medical treatment.

Question 10

WHAT TEST TO USE

Answer 10

INTERLEAVES SLIDE 16-18

Question 11

What is data dredging?

Answer 11

Inflating the Type 1 error rate by carrying out many statistical tests on a single set of data.
- Seriously decreases power of each individual test.

Question 12

Stats mistakes facts:

Answer 12

30%-50% of medical studies include stats mistakes, 8% is enough to change conclusions.

Question 13

What is analytical flexibility?

Answer 13

A number of different statistical methods and decisions could be applied to address the same hypothesis.
This is especially true for complex study designs.
These different methods can yield different conclusions, but hopefully most converge.

Question 14

What is publication bias?

Answer 14

What we see in the peer-reviewed literature is a biased representation of the truth.
Ex. bias toward positive result, HARKing, etc.