regularity stance: correlation is causation interventionist: experiments capacity view: importance of knowledge of the whole system
Statistics
Bayesian Network model: reduce spurious correlation (rule out uncorrelated A-cause-B by observe with variable C)
need supplementary evidence (consisting with existing theory)
problem of under-determination
Disadvantage
Disadvantage
complexity
reflexively
"external validity": context-sensitive - result produced in isolated system can't generalize to larger context
impossible isolating confounding in social science
Two roles
theory testing (empiricist) by observation -> to falsify theory
experimental control (interventionist) -> to isolate and manipulate a purported cause.
Hacking: experiment is a reliable production of phenomena
Purpose: mimic experimentation
Advantage
easy to use
assume identical population
Disadvantage
I believe there is a plausible common cause for each of the three cases which I will justify by enumeration. The condition of "being poor" and the condition of "premature death" can both be caused by "underlying disease." Since "disease" reduces one's ability to work, "disease" can cause someone to "being poor". "Disease" can cause "premature death" of such disease. In the second case, a "good parenting style" can be a common cause of "having pre-school education" and "above-average high school grades" since a good parent demands the child to have "pre-school education" and at the same time spends more time educating the child which will lead to "above-average high school grades." In the third case, "being a developed country" can be a common cause for its "government providing health service" and "longer average life spans" of citizens since a developed country has more resource for "health service." In addition, water in the developed country tend to be clean which lengthens the "average life spans". In all cases, there is a plausible common cause.
approach to justify causality relates to one's ontological stance on causality.
we can only justify thing is the thing has a clear definition
any good justification is to manipulate evidence to fit definition
therefore a justification is closely relates to the definition
since a ontological stance on causality is extensively equivalence to having a stance on the definition of causality, approach to justify causality relates to one's ontological stance on causality.
correlation of "rural area" and "voting for Republican" to causality
I should not conclude causality
because correlation does not automatically infer causality (due to multiple possible diagram of Bayes Net)
one would argue causality is one has supplemental evidence, or if one believes in "regularity stance"
check this out https://www.tylervigen.com/spurious-correlations
potential limitation with experiments?
trading in real word is more complex than in experiments
there are hidden value: friendship, subjective value of good that are in play
if any trading in real world have potential outside variable that is in affect, the experiment is useless
capacity: correlation is result from causation
Table of Content