experimental design Flashcards
What is the scientific experimental design layout?
What are the principles of experimental design?
-Observation:
As cockle density increases, size/growth rate appears lower
-Hypothesis:
Ho = cockle density does not affect growth rate
Ha(lternative) = cockle density affects growth rate
-Methods:
Experimental units (U) = 0.25 m2 plots
Treatments (T) = 0.5, 1, 2 & 4 × natural density
Measurement variable = change in shell height after 6 months
-Design:
Arrangement of U in space (layout)
Number of U per T (number of replicates, n)
Measurement sequence
Statistical analysis
Why is replication necessary in experimental design?
-Provides estimate of the error
Inherent variability (e.g differences between individual growth rates)
Experimenter induced (e.g ability to measure small differences in shell height)
-Increases precision of treatment means
Increased ability to detect smaller differences among treatment means
-Choosing the number of replicates
A function of the precision required to detect a given difference among treatment means
Calculated from previous data, pilot studies (see Kingsford & Battershill)
Most likely a function of financial/logistical constraints
What are the four key elements of experimental design?
Replication
Randomization
Reduction of error
Range of validity
Describe the concept of randomization in experimental design
-Provides an unbiased estimate of the experimental error and the treatment mean error
Critical for statistical tests
-For example:
Location of experimental plots on the shore
Allocation of density treatments to experimental plots
Allocation of cockles to treatments plots
Describe the concept of reducing error in experimental design
-Blocking RCB
-Method
Blocks laid out across gradient
Random assignment 1 replicate T block-1
-Criterion
Identifiable source of variation among U due to position (e.g environmental gradient)
-Advantage
Removes source of variability from error term
Increases sensitivity for detecting T effects
Ensures interspersion
-Limitations
If there is no block effect reduced power
Assumes no interaction between block and T
Describe the concept of increasing complexity in experimental design
- Replicate T within blocks
- Tests
Effect of environmental gradient (compare among shore locations)
Effect of density (compare among density Ts)
Interaction between density & environment (determine whether the effect of the environmental gradient is consistent among density Ts)
Analysis – two way rather than one way ANVOA (see next lecture)
Describe the concept of reducing error in experimental design
-Initial homogeneity of U
Variability amongst U may add to error reducing ability to detect differences among T
e.g removal of non-experimental fauna from plots
Applicability of the results?
-Concomitant measurements
Additional measurements that may help explain variation among replicates within a T
e.g. account for variations in initial cockle shell height (growth rate is often related to size)
Describe the concept of range of validity in experimental design
-Fixed factors
Treatment levels are fixed by the experimenter in order to measure a response
e.g cockle density T were set at 0.5, 1, 2 & 4 × natural density
Scope of inference = levels you have tested
-Random factors
Treatment levels are chosen at random from a range of possible values
e.g measured cockle density in 20 randomly placed quadrats and have randomly chosen 4 for experimental purposes
Scope of inference = factor as a whole
-DO NOT extrapolate beyond the limits of your experiment
what are some of the mistakes people make when designing experiments?
1. Many experiments involve manipulations which can introduce artifacts
2. Cages & fences
- Used to exclude/include different trophic groups
- e.g predation & soft sediment community structure
- Alter microhabitat
Affect flow, shade, create structure
-Alter behavior
-Assessment of artifacts
Partial cages
Direct measurements
Routine monitoring
- Procedural affects may obscure/confound treatment affects
- Alternatives
Chemical barriers
Maintain buffer zones
3. Transplant experiments
-Often used to determine impact of environmental conditions on organism performance
Standardize for differences among populations by transplanting from one site
-e.g effects of sewage outfall cockle growth
-Must include a procedural control for handling effects
4. Laboratory experiments
-Sham operations
-Placebos
What problems in experimental design are there and how to avoid them?
