Causality (bookmark page 104)

Question 1

Correlation vs Causality (with formulae)

Question 2

Examples

Answer 2

Drug –> Reduction of illness
More Ice-cream = More Sharks (temperature)

Question 3

Associational vs Causal Concepts

Answer 3

Associational concept can be defined in terms of joint distribution of variables e.g., correlation, likelihood

Causal concepts cannot be defined in terms of distributions alone. e.g., spuriousness, effect, confouding. Verified through experimental control.

Question 4

Define P(Y=y, do(X=x))

Answer 4

Probability of effect ‘y’ if treatment condition X=x were applied to the population

Question 5

do-operator

Question 6

Exogenous vs Endogenous variables

Answer 6

Exogenous variables: observed or unobserved variables that are kept unexplained in the model i.e., not modeled to be resulting from other variables in the model

Endogenous variables = effect of other exogenous/ endogenous variables

Question 7

Structural Causal Model

Answer 7

Endogenous, Exogenous variables in the model + functions specifying how each endogenous variable is generated from other variables

Symptom := f(Disease, Other Stuff)
Disease := g(genetics, socio-economic level, etc)

Symptom/Disease = Endo
Other Stuff could be modeled to be exo

Question 8

First Order Markov Assumption

Answer 8

New state depends on previous state and states prior to that need not be considered

Question 9

Blocked paths

Answer 9

Chain:
A –> B –> C
blocked path: C conditioned on B is independent of A

Fork:
A –> B
A –> C
Blocked path: B conditioned on A is independent of C
Unblocked path: B (not conditioned on A) is associated with C due to common cause A

Question 10

Bookmark

Answer 10

https://www.youtube.com/watch?v=U1S8Rq8IcrY&list=PLoazKTcS0Rzb6bb9L508cyJ1z-U9iWkA0&index=32