Hypothesis, sampling + design Flashcards
Null Hypothesis
Ho - always a statement of no difference
No statistical significane
Alternative Hypothesis
H1/Ha - leads to change
Statistical significance
Observational Studies
Applicable to real world
Less/no control over experiment
Confounding factors - may lead to wrong conclusions
e.g. animals in the wild
Experimental studies
Full control of factors (diet, environmental)
Less applicable to real world - animals in captivity
Studies
Observational + Experimental
Combination of both give best understanding
Sampling
Can only generalise from samples if randomly collected
Spatial
Random
Timescale
Spatial sampling
Lay a grid over area
Use random number generator to choose GPS co-ordinates of sample squares
Random sampling
Allocate treatment groups to individuals using a tag/code
Random number generator
Eliminates unconscious bias
Timescale sampling
Alternate measurements over time - not confounding
Batch effects
Variation of results in different batches could be due to external factors
eliminate by randomising use of kits
Investigator Bias
Swap factors to avoid subjectivity affecting estimations in experiments
Design
Spatial
Blocking
Latin Square
Control
Spatial design
Equal numbers of treatments in each of the replicate blocks used - not confounded by position
Blocking design
Mitigate gradient effects that could confound
Spread treatments equally across areas - guards against other effects
Don’t reduce spatial variation - prevent it from confounding your experimental factor
Latin square
All varieties present in each row/column
edge effects equal for each variety - equal number of each on edge
Don’t reduce spatial variation - prevent it from confounding your experimental factor
