Abstract: With SPSS software,this paper treated the four indexes used to measure house prices as the factors to affect the sales amount of the completed apartments of commercial houses.A price model was constructed based on the idea of multiple linear regression.It can be seen  that the residual cumulative probability plots are fluctuations around the baseline in the residual sequence of pure random test,the sample autocorrelation coefficient gradually decreases to zero,and basically there is no autocorrelation about residual series.

Key words: multiple regression, goodness of fit, residual analysis, analysis of different variance