Experimental design Flashcards
What is fidelity?
overall proportionate difference (overall similarity of the organisms)
o High fidelity simply means that all properties are equally badly reproduced
o Fidelity becomes more important in the later stages of translational and applied research
What is discrimination?
the extent to which the model reproduces one particular property of the original, in which we happen to be interested
o Allows the use of “lower” vertebrates, invertebrates, ex vivo models, tissue culture, bacteria, etc.
What is bias?
Statistical bias is anything that leads to a systematic difference between the true parameters of a population and the statistics used to estimate those parameters. In other words, bias refers to a flaw in the experiment design or data collection process, which generates results that don’t accurately represent the population.
How can we account for bias / subconcious bias in our experiments?
Randomisation of animals to treatment
Blinding of assay to treatment
Identification of confounding factors
What is a false Positive?
A false positive is the chance that you will see a significant result by chance. Statisticians refer to this as a ‘type I error’ (P is first in the alphabet so will be first type of error)
It is specified by the significance level. You should use a value of p=0.05. Higher levels of
significance will need to be justified, as they require more animals.
What are false negatives?
Every time that we carry out an experiment there is a chance that we won’t see a real effect – we will have a false negative. Statisticians call this a ‘type II error’.
The chance of seeing a real effect as statistically significant is the power of the experiment. Power = 1-false negatives
Power up to at least 0.8 (typically) and better 0.9 or even 0.95
to reduce the chance of false negatives.
What is power?
Power is the probability of rejecting a false null hypothesis. Formally, power is typically given by 1-β. Again, to differentiate power from the treatment effect β, in this resource we will denote power by 1-κ. That is, maximizing statistical power is to minimize the likelihood of committing a type II error.
What is the experimental unit?
The experimental unit is the smallest unit which can
independently respond to a treatment.
It is important to correctly specify this because this is
the replicate.
If housing animals in groups and giving all animals the same treatment the cage is considered the replicate - this is due to the confounding cage effect.
d
What is a confounding factor?
factor applied unequally. Differences
seen in your experiment are due to something other
than your treatment.
What is a nuisance variable?
A nuisance variable, which is of no particular interest to you in itself but needs to be controlled or accounted for in the statistical analysis, so that it does not conceal the effect of a variable of interest.
When to integrate sex into research or not?
Don’t:
when there are sex specific effects e.g ovarian cancer
when there are different models in males and females such as lupus (f) (Lupus develops so late in males that they are basically geriatric and they have a whole heap of health issues from being geriatric that are not related to Lupus )and kidney damage induced hypertension (M)
How to integrate sex into research designs?
research designs that allow sex to be integrated - ANOVA
T-test wont allow you to integrate sex into a study.
What is blocking?
Blocking systematically removes variability thereby increasing power.
Use this only for parameters you are not interested in e.g. sex only when you are NOT interested on comparing males and females. batches of animals, day of harvesting/assay
block of housing, researcher carrying out the duty.
each block should be a mini experiment with at least one replicate of
each treatment combination in every block.
WITHIN a block, keep conditions as homogeneous
as possible at the expense of variation BETWEEN
blocks.
Put as many sources of variability as possible into the
one block e.g. 2 blocks.
Block 1. Researcher1/ batch1/ day1
Block 2 Researcher2/ batch 2/day2
replicates are ideally equally distributed throughout your blocks.
What is a factoring?
Systematically removes variation from error thereby
increasing power
Allows you to quantify
Individual treatment effects e.g. Drug effect, gender effect
Interactions e.g. Drug X gender effects
Use for parameters that you ARE interested in
Treatment!
Gender/ age/ strain?
Have a MAX of 3 factors
what are the advantages of blocking?
Very highly significant effects can be detected early
Unsuccessful treatments can be stopped early
Potential 20% to 30% saving in animals
what are the disadvantages of blocking?
Some kind of ‘Bonferroni’ correction has to be applied which
reduces overall power.
Design can be complicated – may need to consult a statistician
How many controls do you need per block?
at least 1 per block
What are the 3 types of data?
Quantitative (measurable with percision e.g white-blood cell count) us ANOVA and t-tests
Ordinal. Ranked/subjective data e.g. tumour grade - use non-parametric tests e.g. chi-squared test.
qualitative, categorical data. e.g presence/absence of cancer. use qualitative tests e.g.
