0
$\begingroup$

I originally posted this question here. But I think it is better suited for here.

I am running Generalized Method of Moments (GMM) Estimation for Panel Data in R using the plm. I have set of valid instruments, and I would like to select one using BIC-GMM or AIC-GMM criteria for instance. However, when I tried to compute using the standard R command AIC, but shows an error. I would like to compute a function from scratch, but I don't know which is the right formula for the GMM context. I'm using the differenced transformation, which basically is the Arrellano and Bond, (1991) estimator.

library('plm')
plm.gmm <-
  pgmm(
  dynformula(as.formula(model), lg),
  data = pdata,
  effect = 'twoway',
  model = 'twostep',
  transformation = 'd',
  gmm.inst = Z[[i]],
  lag.gmm =  c(lg.gmm[[i]][[1]], lg.gmm[[i]][[2]])
  )
AIC(plm.gmm, k=2)
Error in UseMethod("logLik") : 
  no applicable method for 'logLik' applied to an object of class "c('pgmm', 'panelmodel')"

In this website I found the following formula: $$AIC= N + SSE*\log{(k*\pi)} + N*\log{(SSE/N)} + k * (1+1)$$

  • N=Number of observations.
  • SSE=Sum of squared residuals.
  • k=Penalty per parameter.

I have two questions, can I use this formula in the context of GMM panel estimation, and what it's the intuition behind (1+1) term.

$\endgroup$
  • $\begingroup$ how do you get the plm.gmm object? $\endgroup$ – DataD'oh Sep 13 '17 at 11:18
  • $\begingroup$ What do you mean by GMM? Gaussian mixture models? $\endgroup$ – DataD'oh Sep 13 '17 at 11:19
  • $\begingroup$ Hi – DataD'oh, I changed a bit my question to reply to your comments. $\endgroup$ – Mario GS Sep 13 '17 at 12:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.