[R-sig-ME] Assumptions of random effects for unbiased estimates

[Laura Dee]

Dear all,
Random effects are more efficient estimators – however they come at the
cost of the assumption that the random effect is not correlated with the
included explanatory variables. Otherwise, using random effects leads to
biased estimates (e.g., as laid out in Woolridge
<https://faculty.fuqua.duke.edu/~moorman/Wooldridge,%20FE%20and%20RE.pdf>'s
Econometrics text). This assumption is a strong one for many observational
datasets, and most analyses in economics do not use random effects for this
reason. *Is there a reason why observational ecological datasets would be
fundamentally different that I am missing? Why is this important assumption
(to have unbiased estimates from random effects) not emphasized in ecology?
*

Thanks!

Laura

-- 
Laura Dee
Post-doctoral Associate
University of Minnesota
ledee at umn.edu
lauraedee.com

	[[alternative HTML version deleted]]