causal modeling

Question 1

causation (modeling)

Answer 1

the influence that components of a system exert on one another (“component” is used vs “variable”; component may reflect some underlying quality, like intelligence, which IQ measures but not perfectly)

Question 2

latent variable

Answer 2

unobserved variables, the theoretical concepts or constructs about which we are interested in making inferences (eg intelligence)

Question 3

empirical indicator

Answer 3

the observed variables, the measurements actually obtained in the study (eg IQ score)

Question 4

exogenous variable

Answer 4

• the latent variables whose causes are not represented by the model
• so nothing in the model’s scope “informs” the exogenous variables–in terms of causal diagrams, only arrows “leave” exogenous variables, and no arrows point toward them

Question 5

endogenous variable

Answer 5

• latent variables whose causes are represented within the model
• have at least one variable that has a causal effect on them–ie at least one arrow in the model points toward an endogenous variable
• eg, outside influences to the model (eg changes in government policy)

Question 6

correlating vs non-correlating pathways

Answer 6

• correlating–“there [must be] some node on the pathway from which you can get to both A and B by following the causal flow” (ie “source” node need not be on either end)
• non-correlating–these do not follow the directions of the causal links

Question 7

causal diagrams

Answer 7

• aka hypothetical causal network, reflecting a “state of belief” by the modeler
• nodes:
  
  • square–can be measured
  • round–cannot be measured (aka latent variable)
• links: arrows can be one way (A (tail) causes B (head)), or two-way (signifies a correlation vs causal)