Chapter 7 - Variation, Replication and Randomization Flashcards
Explain Variation:
A change or difference in condition, amount, or level, typically with certain limits.
___________ can be of direct interest to us.
Variation
______________________, or the type of variation we are not interested in, is inevitable.
Random variation
Define Replication:
A way of “cancelling out” the random variation among replicates within a treatment, provided each replicate is truly independent.
Replicates must be _____________-.
independent
Define Randomization:
A way of ensuring each replicate is independent (such that you actually do end up cancelling out the random variation among your replicates).
What are the 4 types of Measures of Variation?
- Range
- Interquartile Range
- Variance
- Standard Deviation
Explain the Interquartile Range of variation measurement:
The range where 50% of the data points fall between.
How do we know when replication is needed in an experiment?
When other factors / things cause random variation in the variable.
What is needed to ensure replication to be successful at cancelling out random variation and isolating effects of the independent variables?
Samples / replicates must be independent of each other.
What does Independence mean?
That every individual has an equal chance of being + / - deviation from the TRUE norm / mean of the treatment group.
When does Psuedoreplication occur?
When the number of measured values or data points exceeds the number of genuine replicates
Provide an example of Psuedoreplication:
Sampling 10 turtles from the same nest might be problematic.
What are the 5 types of Pseudoreplication?
- Common Environment
- Relatedness
- Stimulus
- Behaviour Influence
- Time
What is the Behaviour Influence of Pseudoreplication?
Physical or emotional variables are shared by subjects.
Provide an example of Behaviour Influencing in Pseudoreplication:
A single aggressive individual in a group will make everyone else aggressive.
How does Time incorporate to Pseudoreplication?
Understanding time differences in measurements (E.g., Measuring by hour or minutes).
What are 3 ways to avoid Pseudoreplication?
- Overcome Stimulus bias
- Increase Replication
- Incorporate Randomization
What are the two steps to determine the appropriate number of replicates?
- Educated Guess on what others have done
- Do a Power Analysis
What is a Power Analysis?
Experimental “power” is the probability your experiment will yield the correct result
What 3 factors does a Power Analysis depend on?
- The difference between treatments.
- The degree of random variation within your treatments.
- The number of replicates within treatments.
The bigger the difference; the easier it is to __________________.
detect a difference
More variation makes it ________ to detect a difference.
harder
The more replication, the more likely you are to _____________, if one exists.
detect a difference
What is a Type I Error?
When your experiment erroneously detects a difference when there is none in real life.
What is the Rate of a Type I Error?
The probability that your experiment finds a difference when there is none (aka the p-value).
What is a Type II Error?
There is a difference in real life, but you do not find one.
What is the Rate of Type II Error?
Type probability that your experiment does not find a difference when one actually does exist.
Type II Error = ________________
1 - (Power of Experiment)
As Type I ___________, Type II ____________.
Increases, decreases.
