Experimental design Flashcards
Lecture 8
What are observational experiemnets?
involve making measurements of systems.
- What is the density of birds in a forest fragment?
What are Comparative experiemnts?
measures properties of a system more than one point and ask if there is a difference betweeen them.
- does the decomposition rate of leaves decline with elevation?
What are manipulative experiments?
impose different treatments on different experiemntal units sampled from the population.
Create a new set of circumstances from which we can observe the behaviour of the system and test if predictions are true or not.
how do we define the population?
influences how we sample by defining the spatial boundaries of the study.
ensures that the population we define is the one we sample.
gives us the context and the general ability of the our study.
how can we reliably estimate population parameters?
ensure that our sample allow us to relaiably estimate population parameters.
in manipulative experiments, to determine the “population response” to experimental treatments.
Key principles:
Randomisation
Replication
Control
What is bias?
statistical sampling or testing error caused by systematically favouring some outcomes over others.
What is randomisation?
Avoiding bias.
Need to ensure that samples are representative of the population of interest.
Treatments are assigned to units at random, such as by flipping a coin.
What is replication?
An independent sample drawn from the population of interest.
How does replication increase reliability?
The greater the number of replicates (greater sample size), the more reliable are our estimates of population parameters, the greater the power, and the more chances we have of detecting statistically significant differences.
- Controls for random or stochastic error.
What is a control group?
A control group is a group of subjects who do not receive the treatment of interest but who otherwise experience similar conditions as the treated subjects.
E.g. injecting an animal with a drug may cause changes to the animal independent of any effect of the drug. We would want to include a control, e.g. injecting animals with a saline solution, in order to isolate the effect of the drug from the procedure of administering it.
What is pseudoreplication?
and why should it be avoided?
Subsample = pseudoreplicate
Key issue = violates a key assumption of stats = independence of replicates. The reliability of our parameter estimates will also be inflated (larger sample sizes mean smaller standard errors), so we are more likely to detect significant differences where none really exist. That is, we are more likely to make a Type I error (reject H0 when H0 is true).
How can we control psuedoreplication?
Plan your study thoroughly:
- Clearly define your independent and dependent variables.
- What constitutes a unit of data?
- What is the unit of replication?
- Randomise your sample sign.
