**3. Research Methods**

#### *3.1. Methodology*

Regarding research methodology, this paper adopts a quantile regression model and the common ordinary least squares (OLS) linear regression model found in the literature. Compared with an OLS linear regression model based on mean, a quantile regression model has two advantages. First, the quantile is less likely to be affected by extreme values than the mean. Although such a problem can be alleviated by taking logarithms of variables in the mean regression, quantile regression is better in terms of managing skewed variables, such as income and housing value. Second, regression equations can be established according to different quantile levels of dependent variables. Therefore, all factors can be closely studied under the influence of all sorts of housing profitability levels. Furthermore, the changing patterns of these factors in different quantile levels can be analyzed, which is better than the over-general conclusion reached under the guidance of the OLS model.

This paper adopts 19 quantiles from 5 to 95 at an interval of 5; additionally, it selects a multivariable linear regression model, in which every dependent variable has 20 regression equations. Thus, an analysis based on regression excels can be conducted as performed in classic quantitative research. Additionally, regression coefficient tables in different quantiles can be created to inspect the changing patterns of independent variables in different quantile levels.
