Patent US 11,288,690

A method for estimating a real estate price index based on a repeat sales model. The method includes the steps of eliminating corrupt portions of data used to determine price indices; regressing a difference in a log price of a second sale minus a log price of a first sale for a property on a set of dummy variables, eliminating the observations with zero residuals for estimating the dispersion function; using a quadratic formula to model a dispersion of residual values with respect to time interval; using the quadratic formula to solve for a dispersion for each time interval starting with a minimum positive value; determining a minimum dispersion value that is a value calculated from the quadratic formula; determining a maximum dispersion value that is calculated from the quadratic formula for each time interval starting with a minimum positive value, wherein the dispersion values are prevented from decreasing with increasing time intervals; repeating the step of regressing using the inverse of the square root of fitted values from the quadratic formula as weights; identifying and selecting the time periods and associated data for estimating a price index that covers a particular time period; excluding the index values which are determined to be unreliable; and obtaining an index series for a lower geographic level by using the index series of a higher geographic level.