I have run an IV model using the plm-package in r. The underlying data is panel data.

My instrument is “weak”, i.e. the F-statistic lies below 10.
How can I implement an Anderson-Rubin test for my instrument and calculate a confidence interval? I’ve read that this kind of test does not require instrument relevance.

fe_model <- plm(
  log(Y) ~ log(X) + Controls
  |  log(instrument)  + Controls,
  model = "within",
  data = data,
  index = c("country","year"),
  effect = "twoways")

# HAC standard errors
fe_model_se <- coeftest(fe_model, vcov = vcovHC(fe_model, type = "HC3"))[, 2]