---

The main objective of this research paper is to select an appropriate model for time series forecasting of total import of Bangladesh. The decision through out this study is mainly concerned with autoregressive integrated moving average(ARIMA) model, holt-winters´ trend and seasonality model with seasonality modeled additively and vector autoregressive(VAR) model with some other relevant variables. In this research the analysis has been done on a set of data based on total import of Bangladesh during the period July 1998 to July 2009. Here an approach is made to derive a unique and suitable forecasting model of total import of Bangladesh. From the study it is found that vector autoregressive model of total import of Bangladesh, where, total export of Bangladesh and net foreign asset of Bangladesh are taken as other endogenous variables, gives us less forecasting error than that of others. So, it is proposed that for forecasting total import of Bangladesh one can use this VAR model. But before using this model one must verify the validation of the model in different time periods, because a forecasting model may not remain valid and suitable as time changes.

---

Tanvir Khan has graduated in Applied Statistics from University of Dhaka. He is now a student of M.S. and hoping to obtain this degree in May 2011. His main research interests are econometrics, modeling time series data, factor analysis and statistical computing. He is a regular writer about statistical programming in www.stattler.com.

---