The Estimation and Interpretation of Urban Density Gradients
通过引入Box-Cox变换和异方差误差，推广了负指数变参数密度梯度模型，并基于人口密度期望值对模型条件变量的变化响应进行解释，利用1970年美国39个城市的数据进行实证分析。
The negative-exponential varying-parameter, density-gradient model is generalized by introducing the Box–Cox transformation in conjunction with truncated and heteroscedastic errors. An interpretation of this density-gradient model is presented, based directly on the magnitude and response of the expected value of population density to changes in the model's conditioning variables. Previous studies interpreted the gradient parameters purely as descriptive statistics of decentralization and mistakenly trivialized direct inference on density levels by implicitly interpreting the gradient framework as purely deterministic. For empirical analysis, we used data on 43 randomly selected census tracts in each of 39 U.S. cities in 1970.