The invention provides a statistical modeling method for measuring and calculating a carbon emission scale, which is characterized in that the conditions that data is from a heterogeneous population and the data has deflection are considered, and which factor and the influence degree of carbon emission are researched. Comprising the following steps: step 1, collecting carbon emission data of a plurality of regions; 2, analyzing factors influencing the carbon emission scale, and searching related data about novel urbanization; 3, constructing a novel urbanization evaluation system to define the urbanization level; 4, estimating the carbon emission of the region by using an emission factor method; step 5, analyzing carbon emission data of each area by using a Tapio unhooking model, and finding out individual differences and generality; 6, studying factors influencing the carbon emission scale by adopting an STIRPAT model, and reducing the multicollinearity by using a principal component analysis method; and 7, measuring and calculating the carbon emission data by using a hybrid expert regression model.