I am researching the effect of minimum wage (logged) on labor force participation (binary), using individual level data. I've got about 16million observations. I'm using Stata 15.1. For a few reasons, I'm using an LPM, not logit/probit.
I created a categorial variable 'educ': 1=not finished HS, 2=high school grad, 3=some college, and 4=college grad.
My actual regression is far more complicated because I add a bunch of covariates, fixed effects, a time trend, clustered standard errors, etc.
But the parts that are important for the current issue I'm having are as follows:
Code:
reg ilf i.educ#c.lmwage i.educ
Then I run code to identify the exact name Stata gives each of the interacted factor variables, and finally, run coefplot.
Code:
regress, coefflegend nolabel coefplot, keep (1b.educ#*.lmwage 2.educ#*.lmwage 3.educ#*.lmwage 4.educ#*.lmwage)
My only solution was to hard code four interaction terms then run the following work-around:
Code:
reg Ilf HSdroplmwage HSgradlmwage SomeCollmwage Collgradlmwage i.educ coefplot, keep (HSdroplmwage HSgradlmwage somecollmwage colgradlmwage) vertical yline(0)
Any thoughts?
0 Response to Coefplot omitting a factor variable interaction coefficient
Post a Comment