Differences between revisions 8 and 9
Revision 8 as of 2025-04-29 19:24:20
Size: 1061
Comment: Relabel
Revision 9 as of 2025-05-17 03:49:08
Size: 0
Comment:
Deletions are marked like this. Additions are marked like this.
Line 1: Line 1:
= Exogeneity =

'''Exogeneity''' means that all predictors are independent of the outcome and the error term.

<<TableOfContents>>

----



== Description ==

Mathematically, exogeneity and [[Statistics/Homoskedasticity|homoskedasticity]] together can be expressed as:

{{attachment:exo.svg}}

The full set of variables is independently and normally distributed about 0. The [[Statistics/CovarianceMatrices|covariance matrix]] is fully specified as the [[LinearAlgebra/SpecialMatrices#Diagonal_Matrices|diagonal matrix]] of variances.

The opposite condition is '''endogeneity'''.

An assumption of exogeneity would be violated if...

 * confounding variables were omitted
 * in time series models, a lagged dependent variable can be correlated to the error term
 * if OLS is mistakenly used on a system of equations (i.e. '''simultaneous equation bias''')

It can also be useful to express exogeneity as a [[Statistics/ConditionalExpectations|conditional expectation]] like:

{{attachment:cond.svg}}



----
CategoryRicottone