endogeneity

Question 1

endogeneity (theoretical definition)

Answer 1

when marginal distribution of predictor variable is not independent of the conditional distribution of outcome variable given the predictor variable

Question 2

marginal distribution

Answer 2

events value independent of other events

Question 3

conditional distribution

Answer 3

events value given value of another event

Question 4

endogeneity (practical definition)

Answer 4

when predictor variable x is correlated with error term (e)

Question 5

what does endogeneity cause

Answer 5

bias in B estimates
cannot test if variables are endogenous
if endogeneity is identified, must determine why it exists to solve the problem

Question 6

sources of endogeneity

Answer 6

simultaneity bias
omitted/confounding variables
measurement error

Question 7

simultaneity bias and solution

Answer 7

when causality goes both ways
x causes y which causes x
solution = use stats method developed for this (2-stage least squares regression)

Question 8

omitted / confounding variables and solution

Answer 8

when x is correlated with both outcome and omitted variables z, variance explained by z falls on error
solution = ensure potential confounds are measured and included in the model

Question 9

measurement error and solution

Answer 9

measure x* instead of x
x* = measurement of x with error included (r) (measurement error becomes part of e but is associated with x)
solution = careful planning and study design and including pilot testing