The paper studies regression analysis for interval data which are obtained by packaging point data． The existing methods of regression analysis for interval data only consider the endpoints of the intervals． This study focuses on how to use the point data within intervals sufficiently． Firstly，when the error of the regression model for point data satisfies the supposed three characteristics，the error of the regression model for interval data also follows the three characteristics． Then，based on the descriptive statistics of interval symbolic data，a novel approach of estimating the parameters of the regression model is proposed． A method is given to predict the interval values of the dependent variables． Finally，a simulation experiment and an application case are performed following the proposed methodology． The results indicate a better performance of our method than the existing CCＲM method．