Chapter 12 - Advanced Design Ideas Flashcards
How should you decide how to sample?
Ultimately, design your sampling strategy to capture variation and avoid Pseudoreplication.
What are the four types of sampling?
- Sequential Sampling
- Stratified Sampling
- Systematic Sampling
- Interactions (2-way)
Define Sequential Sampling:
Sample until you’ve satisfactory captured the variation among individuals in your treatment group (you’re not getting any new / anomalous values anymore)
When should you use Sequential Sampling?
Only works if you can input / analyze your data “as you go”, but prevents you from taking more samples than you need!
Define Stratified Sampling:
Divide study areas into different types ahead of time (E.g., fields and woods), then randomly pick locations to sample within each type such that sample numbers in each type match their proportions on the landscape.
What is a limitation of Stratified Sampling?
It doesn’t necessarily allow you to analyze differences between types, it just ensures you’re representing all the types.
Define Systematic Sampling:
Generally a random approach to sampling is best, but systematic sampling can be appropriate depending on the question (E.g., transects across gradients).
What does Interaction sampling assess?
Assess the effects of two or more independent variables (and their interactions) at once.
When is it helpful to use Interaction sampling?
When you know the effect of one factor depends on the level of another, but decreases your statistical power because you’re chopping your sample size (per treatment) in half.
When does statistical power increase?
With the size of the replication so by choosing two extreme treatments, you’re maximizing statistical power.
How can you get the best estimate of the shape of a relationship in your design?
Need to have as many treatment groups as possible, spanning the range that you are interested in, and as evenly spaced as possible.
Define Subsampling:
Deciding how to allocate your sampling efforts between two different levels.
When is Sequential Sampling only practical?
If measurements can be extracted from a single sample easily, before the next sample is taken.
When do you stop sampling when using sequential sampling?
When the ongoing analysis indicates that we have collected enough data to answer the question at hand.
What is another work for continuous variables?
Covariates
