Tests of Fit Using Sample Moments
研究了Gurland & Dahiya(1970)提出的基于矩和广义最小二乘的拟合优度检验方法，给出了正态、指数、逆高斯和伽马分布下检验统计量的分量表达式，并讨论了参数冗余和稳定性性质。
A method for constructing tests of goodness-of-fit using moments and generalized least squares was given by Gurland & Dahiya (1970). First, it is shown that the method gives statistics that can be expressed as a sum of components; expressions are found for the components in the normal, exponential, inverse Gaussian and gamma cases. Second, a result is obtained which shows that in some distributions certain parameters are redundant when a test of fit is required; this result leads to a simplification in the calculation of the test statistic. Third, in some cases the parameter estimates in the least squares procedure exhibit a stability property and conditions are given for this to occur. All three of these properties are interpreted in terms of regression models.