Planning, designing, performing and publishing animal experiments Flashcards
(10 cards)
1
Q
Good experimental design will help me to
A
- improve the quality of my scientific work
- get published in better journals
- save time and money
- use fewer animals
2
Q
On which (six) variables does the power analysis depend?
A
- group size
- relevant effect of treatment (signal)
- standard deviation (noise)
- desired power
- level of significance
- alternative hypothesis
3
Q
Types of experiments:
A
- pilot study
- exploratory experiments
- confirmatory experiments
- experiments to explore relationships between variables
4
Q
Objectives of pilot studies
A
- logistics - setup of experimental protocol
- to analyse effect of treatment
- to analyse variation (noise)
- determine size of experiment
- staff training
5
Q
Objectives of exploratory experiments
A
- develop hypotheses for further testing
- measure many outcomes
- data snoopign –> hypothesis testing
- should be confirmed in a controlled expeirment
- analysed in a ANOVA
6
Q
Objectives of confirmatory experiments
A
- is a comparison
- tests a null hypothesis
- subjects are randomised
- should, if possible, be blind
- need to be powerful - have high probability of detecting an effect
7
Q
Objectives of experiments to explore relationship between variables
A
- tests whether two variables are associated
- if so, the nature of that relationship
8
Q
Name three ways of randomisation
A
- completely randomised design (computer)
- randomised block design
- latin square design
9
Q
Characteristics of randomised block design
A
- deals with known (suspected) heterogeneity
- breaks up an experiment into mini experiments
- subsets of experimental units (animals) e.g. males and females, large and small etc.
(so that you don’t end up having all large or femal animals in one group)
10
Q
What are the properties of a well-planned experiment?
A
- have a clear specification of the aims of the experiment
- is unbiased
- is powerful
- has a wide range of applicability
- should be simple
- should be possible to statistically analyse