In raw reeling silk industry, property analysis and model characterizations of cocoonfilament is an important topic. Cocoon filament time series are different from the othertime series. The main features of such size series are that they are non-stationary and withshort length. In practical research, collecting a large number of such size series can be veryexpensive. So, it is of great value to generate more such size series with the required marginaldistributions and the required autocorrelation structure.In this paper we propose a novel simulation method which can be applied to any non-stationary time series of finite length. Compared with other simulation methods, it guaranteesthat not only the marginal distributions but also the autocorrelation structures of observedand simulated time series are the same. Extensive simulation cases and applications to realcocoon filament time series show that the developed method works very well. |