FK83
commented
8 years ago @BruceZhaoR, thanks for your comment! nf sets the number of future periods for which bvar.sv.tvp computes forecasts. By contrast, the t input to impulse.responses is the past period for which impulse responses are computed. Since bvar.sv.tvp needs a training sample (with length tau, default is 40) as well as p values for the lags (default is p = 1), you can compute impulse responses for 60 - 41 = 19 periods, as you noted in your code. Hence, entering t = 1 in impulse.responses produces results for the 42nd data point (= 42nd row of matrix td), and entering t = 19 produces results for the 60th row of td. Agreed that this handling is perhaps not very intuitive - I'll think about an alternative, perhaps involving labels for the time points (e.g., one label for each row of td).
BruceZhaoR
@FK83 , Thank you for your great job!
I have 60(rows)* 2 (cols) mts matrix, I want to set
nf =20,but it seems doesn't work?here is the snapshot:
I appreciate it if you can give me some instructions, Thank you again!