I am looking to simulate the whole cross-section of daily return series for 20 to 60 days. The purpose is to test some risk measures based no maximum drawdown. Thus, it needs the whole time series.
For non-parametric approaches block-bootstrap seems to work well but for a parametric approach, such as multivariate GARCH, takes too long to fit and I have not found a multivariate time series method that fit in reasonable time to the cross-sectional return daily return data. Any suggestions?