Week 6: R & Experimental Design

Question 1

What three key points do you need to consider in experimental design

Answer 1

1 Cost
2 Time
For experiments
For analysis (complicated set up will take much longer to analyse it)
You may have deadline, BSc, PhD, drug to market
3 Ethics – drugs on humans, or you have to kill mice, or test vaccines or a primate, you don’t want to harm anything unnecessarily

Question 2

You need to consider what the best ways to do things are, if a conservation perspective what is the best way to do this…

Answer 2

If you work for the food and drug regulation, you want to be sure the experiments trying to show that a particular drug works you need to ask the following questions
Are the results real?
Allow clear interpretation? – or are there other interpretation
Can you trust the data?

Question 3

Four Main potential problems

Answer 3

1 Insufficient data/power
• Then trying to draw conclusions based on trends, without actually observing a statistically significant outcome
• You end up over interpreting

2 Pseudoreplication
• Many measurements might be taken but they are not independent measurements, and the same or similar individuals are being measured every time
Or
• The amount of replication of truly independent replication, that the statistical test supposes in the data analysis, then that gives results that are not significant even if the result appears to be

3 Confound factors
• If you see a particular treatment works better somewhere else, different lab, country, you don’t know if it works better because of the different equipment
4 Inappropriate statistics
• Different statistical tests have different underlying predictions
• (we don’t have time to go into the statistics now)
• But statistical tests have very specific assumptions
• The statistics that are used need to be chosen very wisely
• Inaccurate & misleading  wrong
• Some people may be so desperate to publish something might end up going through several stats tests to get a significant p-value

Question 4

Three reasons why problems occur?

Answer 4

They come about due to;

1 Inappropriate design
2 Inappropriate implementation
3 Inappropriate analysis

Question 5

Taking measurements

How do you calibrate measuring instruments (including human observers)?

Answer 5

Steps to reduce 
•	Subjective decision making? 
•	Inter-observer variability?  
•	Intra-observer variability? 
•	Unusable/illegible measurements/notes 
Automation? 
Avoid floor & ceiling effects 
Ensuring that subject are in “natural” conditions 
	Do all that you can to ensure your design is robust

Question 6

Overall

Answer 6

Avoid easy mistakes
Design & statistics are closely interlinked
Consider biology carefully
Better to spend more time planning
Recommended reading “experimental design for the life sciences” third edition Nick Colegrave