WEEK 5: COFOUNDING, EFFECT MODIFICATION Flashcards
Threats to validity
chance, bias (selection, information), cofounding
what is cofounding
distortion of the actual association due to a mixing of effects between the exposure and an incidental variable(s), known as a cofounder
why does cofounding occur
- Because the exposure ground and the unexposed ground are not exchangeable
- They differ by factors other than their exposure status
- Cofounders are particularly a problem for observational studies
- Randomization in experimental studies removes the possibility of cofounding by making the exposed and unexposed ground the same except for the exposure status
criteria for defining a cofounding variable
- Casually associated with the outcome (a true risk factor)
- Noncausal or casually associated with the exposure
- Not an intermediate (mediator) in the casual pathway between exposure and outcome
5 steps to assessing cofounding
- conceptualize the relationship
- think of variables with criteria
- exclude those on the pathway
- identify true cofounders
- deal with cofounders
dealing with cofounding at the design stage
- Restriction
o Limit study inclusion criteria with respect to cofounding factors; study only men or women - Matching
o Produce case and control (exposed/non exposed) groups that have similar characteristics - Randomization
o Experimental studies
dealing with cofounding at the analysis stage
- Standardization
o Age standardization is in fact “adjustment” for age - Stratified analysis
- Include cofounding factors ina. Multivariate regression model
effect modification
*the 3rd factor, called the effect modifier, modifies the effects of the exposure on the outcome
o not a cause of the exposure and outcome ( that’s cofounding)
o not on the casual pathway (intermediary)
cofounding is a problem for________
validity
Mediator
- An intermediary variable occurs between an exposure and an outcome
- Do not need to adjust for the effects of intermediaries in assessing exposure – disease (E – D) associations
what is interaction
when the incidence rate of disease in the presence of 2 or more risk factors differs from the incidence rate expected to result from their individual factors
positive and negative interaction
The effect can be greater than what is expected = positive interaction; synergism
The effect can be less than what is expected = negative interaction; antagonism
what is the best way to reduce sampling error
o increase the rise of the population/study sample
p value meaning
If the p-value < 5%/ probability that results arose by chance <5% (type I error, ) results are considered statistically significant
Interpreting P-Values & Confidence Intervals
If p< 0.05, this does not guarantee the association is real or if p> 0.05, does not mean there is no association; only means that there is not enough evidence to conclude if there is an association
what tools can help assess results if association exists or not
hypothesis tests and p values
When a significance level of 0.05 is used - what does this mean
there is only a 1 in 20 chance that the null hypothesis will be rejected in error (i.e you will conclude there is an association when there is none, the chance of this happening increases as more tests are done)
what does an RR mean for significance
A RR > 2,0 would be considered strong, thus practically significant
An RR < 2.0 wild be dismissed immedisately bc a modest relative risk can lead to high absolute or attributable risk
