Table 2.

The non-parametric stationary block bootstrapping algorithm with a VAR model.

Step 0	Estimate a VAR model of order p using the original data and obtain the point forecast(s).
Step 1	Draw random samples with the stationary resampling procedure from the original data with the same size, estimate the VAR model of order p and obtain the point forecast(s).
Step 2	Repeat step 1 for b times.
Step 3	Combine the b + 1 forecast(s).
Step 4	Calculate the point forecast(s) using mean if the bootstrap distribution is symmetric and median if the bootstrap distribution is skewed. Construct the 95% CI by taking the 2.5th percentile and 97.5th percentile together.