Design of experiments Flashcards
Controlled experiment
- we control the effects of other variable on the treatment /what we are testing
- investigators allocate subjects into different groups
Confounding
occurs when the effect of one variable (X) on another variable (Y) is clouded by the influence of another variable (Z)
Bias
the quantity of interest is systematically under/overestimated, usually caused by confounding variables
Selection bias
If the treatment group is not comparable to control group, then the difference between the two groups can confound the effect of the treatment
Randomised Controlled Trial (RCT)
solution to selection bias - randomises groups
Observer bias
Either subjects or investigator are aware of the identity of the two groups, we can get bias either in the responses of evaluation as they may deliberately or subconsciously resort more or less favourable results
Placebo
“pretending a treatment”, neutral and indistinguishable from the treatment
Placebo affect
subjects thinking they have had the treatment (or even the investigators)
Randomised Controlled Double-Blind Trial
Subjects and investigators are not aware of the identity of the two groups
Placebo needs to be as close as possible to real treatment
Consent bias
When subjects choose whether or not they take part in the experiment
Raises ethical issues
Why is the design of a statistical study important?
it is critical in order to obtain results that can be generalised
Why should we known the design of an experiment when looking at data?
helps us make sure we know what we are testing and if we achieved this - validity
Domain knowledge
background context information that helps you understand the data
If data scientist specialise in one field, may become a domain expert
Reproducible research
requires data sets and software to be made available for verifying published findings and conducting alternative analyses
Meta analysis
collets summarise and draws conclusions from multiple scientific studies on a specific research question
