Calculation of AIC criteria in VAR model
Posted: Mon Oct 31, 2016 12:22 am
Hello,
I found formulas to calculate AIC for vector autoregression model, that show similar results to Eviews calculation for each timeseries separatly in model (ie A1, A2, A3, etc):
Log likelihood = -(n / 2) * Ln(SumOfSquares / n) — (n / 2) * Ln(2 * Pi) — (n / 2)
AIC = 2 * (k / n) — 2 * LogLik / n
n- number of observation, k = 1 — Intercept — trend + lag
But I cannot understand how to calculate AIC for whole model (results that Eviews shows in the end of Vector Autoregression Estimates). Need it to calculate VAR AIC in excel.
Can someone help and explain it?
I found formulas to calculate AIC for vector autoregression model, that show similar results to Eviews calculation for each timeseries separatly in model (ie A1, A2, A3, etc):
Log likelihood = -(n / 2) * Ln(SumOfSquares / n) — (n / 2) * Ln(2 * Pi) — (n / 2)
AIC = 2 * (k / n) — 2 * LogLik / n
n- number of observation, k = 1 — Intercept — trend + lag
But I cannot understand how to calculate AIC for whole model (results that Eviews shows in the end of Vector Autoregression Estimates). Need it to calculate VAR AIC in excel.
Can someone help and explain it?