This paper discusses selected issues related to the use of the multiple linear regression in the real estate valuation. They are the functional form of the model, number of the attributes, trend, scale of the values of the attributes, the form of adjustments, and the issue of the verification of the model. The test based on the sum of the squared error, which allows the assessment of the quality of model is presented. The mean advantage of this test is its simplicity and non-restrictive conditions imposed on the data. Arguments that the dates of sales should be excluded from the list of real estate attributes are brought forward. Also, the paper argues that the regression model should contain an intercept parameter and the scale of the values of the attributes should start from zero.

59-66