Altair slc using instrumental variables to reduce bias in regression
Too long to post to listserv, see github
github (I use UltraEdit editor)
https://github.com/rogerjdeangelis/utl-altair-slc-using-instrumental-variables-to-reduce-bias-in-regression
Bias is a serious problem in regression.
The known validated model for veteran earnings is:
earnings <- 25000 - 5000military + 8000ability
Military service is associated with a loss of $5,000 in earnings per year.
However, we do not have the ability variable.
Without the ability and no instrumental variable, we get
earnings <- 23644 - 1520*military
Military service is associated with loss of $1,520 in earnings per year.
Adding a draft lottery number as an instrumental variable.
The new model
earnings <- 25236 - 4814*military (adjusting for instrumental variable)
Much closer to the known model.
Military service is associated with loss of $4,814in earnings per year.
Input variables
draft_number is the instrument (1-365, lower numbers more likely to be drafted)
military is the endogenous-served variable (1 if served, 0 otherwise)
earnings is the outcome variable, yearly income
related github
https://github.com/rogerjdeangelis/utl-using-instrumental-variables-to-more-precisely-predict-the-income-of-vietnam-veterans
AI query
Please provide a simple real-world example using Vietnam draft lottery numbers as an instrumental variable in regression using r AER package with output
