In my regression, I want to instrument a variable pastS with lgrade and run the following code in R:
modiv <- ivreg(servS ~ horS + size + henley + lpd + tri + turn + rail +
enc20 + wheat | pastS | lgrade, data = regdata)
I get an error: length(formula)[2] %in% 1:2 is not TRUE. Apparently, something is not right with my syntax, but I cannot find an error. Could you help?
AER::ivreg wants you to provide an instrument for each regressor after one separator |, so instead of
> library(AER)
> ivreg(log(packs) ~ population + tax | log(rprice) | rtdiff, data=c1995)
Error in ivreg(log(packs) ~ population + tax | log(rprice) | rtdiff, data = c1995) :
length(formula)[2] %in% 1:2 is not TRUE
it's
> ivreg(log(packs) ~ log(rprice) + population + tax | rtdiff + population + tax, data=c1995)
Call:
ivreg(formula = log(packs) ~ log(rprice) + population + tax | rtdiff + population + tax, data = c1995)
Coefficients:
(Intercept) log(rprice) population tax
7.241e+00 -5.089e-01 -7.907e-09 -4.202e-03
You could consider lfe::felm instead which is more concise and might fit more to your intuition.
> library(lfe)
> felm(log(packs) ~ population + tax | 0 | (log(rprice) ~ rtdiff), data=c1995)
(Intercept) population tax `log(rprice)(fit)`
7.241e+00 -7.907e-09 -4.202e-03 -5.089e-01
Data:
> data("CigarettesSW", package="AER")
> CigarettesSW <- transform(CigarettesSW,
+ rprice=price/cpi,
+ rtdiff=(taxs - tax)/cpi
+ )
> c1995 <- subset(CigarettesSW, year == "1995")