producing data Flashcards
explanatory variables
explain or cause a change in another variable and may be manipulated or have a set value
response variable
the measyred variable that changes in response to the explanatory variable
confounding variables
thought or known to influence both explanatory and response variables - confuses (known beforehand)
lurking variables
like confounding but not known beforehand
simpson’s paradox
unfair averaging
better to compare % at each level
overall % can be misleading
arises because of lurking/confounding variables
observational experiments
- no manipulation of explanatory variable
- no random assignment
- observe what is happening in natural settings (not contrived)
- causation is difficult to establish
- case-control studies
experiments
- manipulation and response
- experimental units (called subjects if human)
- treatments (specific conditions)
- explanatory variables are called factors and have different levels
- factorial experiments look at joint effect of several factors
- placebos and controls
- cause and effect
controls
control lurking variables by comparing treatments
- needed for validity
- keep conditions the same between groups
- use placebo and remove placebo effect
- blinding and double blind
randomisation
use of impersonal chance to assign experimental units to treatments
- matched pairs but chance is better
- each group has similar background and lurking variables
- random allocation
replication
- replicate experiment o many units to reduce chance variation in results
- random error is reduced
blocking
- blocking of treatment groups into sub-sections that are similar with respect to confounding variables
- form of control - sort units into blocks (men and women) first and then randomly allocate
a standard experiment should be
randomised, controlled, double blind and blocked
