It can be very challenging to assess what information to believe. One simple way to evaluate evidence is via something called "Hill's Criteria", considerations to help assess whether an observation has a *causal* component
I'll describe them using xkcd comics
1/12
I'll describe them using xkcd comics
1/12
Sir Austin Bradford Hill, a statistician & epidemiologist, created a list of guidelines for evaluating whether there is evidence of a causal relationship. He determined the following aspects ought to be considered when assessing causality
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1898525/pdf/procrsmed00196-0010.pdf
2/12
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1898525/pdf/procrsmed00196-0010.pdf
2/12
Strength
How big is the effect you are seeing?
Note: Hill suggests that huge effects can suggest causality, however this does not mean small effects cannot
https://xkcd.com/539/
3/12
How big is the effect you are seeing?
Note: Hill suggests that huge effects can suggest causality, however this does not mean small effects cannot
https://xkcd.com/539/
3/12
Consistency
This is essentially reproducibility & replicability
Can your analysis be reproduced?
Has anyone been able to replicate your findings?
http://xkcd.com/242/
4/12
This is essentially reproducibility & replicability
Can your analysis be reproduced?
Has anyone been able to replicate your findings?
http://xkcd.com/242/
4/12
Specificity
Can the association be pinpointed to a specific cause with no other plausible explanation?
http://xkcd.com/1217/
5/12
Can the association be pinpointed to a specific cause with no other plausible explanation?
http://xkcd.com/1217/
5/12
Temporality
Does the timeline make sense?
In general, the exposure ought to come before the outcome it is said to cause.
http://xkcd.com/925/
6/12
Does the timeline make sense?
In general, the exposure ought to come before the outcome it is said to cause.
http://xkcd.com/925/
6/12
Biological gradient
The wording of this point makes it a bit difficult to untangle from the medical application, but generally this refers to a dose effect
Does increasing an exposure yield a change in the outcome.
http://xkcd.com/323/
7/12
The wording of this point makes it a bit difficult to untangle from the medical application, but generally this refers to a dose effect
Does increasing an exposure yield a change in the outcome.
http://xkcd.com/323/
7/12
Plausibility
Does the causal relationship make sense?
This is also a tricky one since plausibility depends on knowledge at the time. If we found it perfectly plausible, we may not need statistics to show the relationship.
http://xkcd.com/605/
8/12
Does the causal relationship make sense?
This is also a tricky one since plausibility depends on knowledge at the time. If we found it perfectly plausible, we may not need statistics to show the relationship.
http://xkcd.com/605/
8/12
Coherence
Similar to plausibility, is there a logical argument that can be made by/to experts in the field regarding causality.
Does it fit into the understanding of the field
http://xkcd.com/1170/
9/12
Similar to plausibility, is there a logical argument that can be made by/to experts in the field regarding causality.
Does it fit into the understanding of the field
http://xkcd.com/1170/
9/12
Experiment
If a controlled experiment can take place, this can strengthen the argument for causality
I view this as a general attempt to implement a counterfactual analysis.
http://xkcd.com/1462/
10/12
If a controlled experiment can take place, this can strengthen the argument for causality
I view this as a general attempt to implement a counterfactual analysis.
http://xkcd.com/1462/
10/12
There is lots of incoming information and data these days, hopefully these principles can help a bit with sifting through the noise! I have a blogpost version of this here if that is your cup of tea:
https://livefreeordichotomize.com/2016/12/15/hill-for-the-data-scientist-an-xkcd-story/
12/12
https://livefreeordichotomize.com/2016/12/15/hill-for-the-data-scientist-an-xkcd-story/
12/12