A Gentle Introduction into SCM
Repo for essay on structural causal models
Association-based Concepts |
Causal Concepts |
Correlation |
Randomization |
Regression |
Confounding |
Conditional Independence |
Disturbance |
Likelihood |
Error Terms |
Odds Ratio |
Structural Coefficients |
Propensity Score |
Spurious Correlation |
Table: Concepts in Causality and Association concepts
Method |
CBN |
SCM |
Prediction |
$\boldsymbol{\cdot}$ Unstable |
$\boldsymbol{\cdot}$ Stable |
|
$\boldsymbol{\cdot}$ Volatile to parameter changes |
$\boldsymbol{\cdot}$ More Natural Specification |
|
$\boldsymbol{\cdot}$ Re-Estimate entire model |
$\boldsymbol{\cdot}$ Only estimate $\Delta$ CM |
|
|
|
Intervention |
$\boldsymbol{\cdot}$ Costly for Non-Markovian Models |
$\boldsymbol{\cdot}$ Pot. Cyclic Representation |
|
$\boldsymbol{\cdot}$ Unstable(Nature CP) |
$\boldsymbol{\cdot}$ Stable(Nature Eq.) |
|
$\boldsymbol{\cdot}$ Only generic estimates($\Delta$ CP) |
$\boldsymbol{\cdot}$ Context specific(Invariance of Eq.) |
|
|
|
Counterfactuals |
$\boldsymbol{\cdot}$ Impossible
|
$\boldsymbol{\cdot}$ Possible |
|
$\boldsymbol{\cdot}$ no information on latent factors($\epsilon$) |
$\boldsymbol{\cdot}$ Inclusion of latent factors |
Causal Inference and Time