Monte Carlo Experiments
Monte Carlo Experiments
复制标题
蒙特卡罗实验
DOI：
--
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Kurt Schmidheiny
中科院分区：
文献类型：
--
作者：
Kurt Schmidheiny
做计量经济学意味着估计参数，如人口的平均值，线性回归中的系数或时间序列的自相关性，给定真实的世界数据的样本。除了点估计本身，我们还想知道我们的估计值与真实值的接近程度。换句话说，我们想知道它的“准确性”或“精确度”。估计量是随机变量的函数（可能很复杂），因此它本身也是随机变量。估计量的性质完全由其概率分布（所谓的抽样分布）描述。然后，可以使用抽样分布来执行针对假设的检验。通常我们对抽样分布的某些矩特别感兴趣，例如均值和方差。在某些情况下，可以从计量经济学模型计算抽样分布。但有时，特别是对于有限（小）样本，这是不可能的或非常困难的。在这些情况下，蒙特卡罗实验是一种直观的方法来获得有关抽样分布的信息，从而获得有关估计量的“质量”。
Doing econometrics means estimating parameters, such as the mean of a population, the coefficients in a linear regression or the autocorrelation of a time series, given a sample of real world data. Besides the point estimate itself, we would like to know how close our estimate is to the true value. In other words we would like to know its “accuracy” or “precision”. An estimator is a (maybe complicated) function of random variables and therefore itself a random variable. The properties of an estimator are fully described by its probability distribution (the so-called sampling distribution). The sampling-distribution can then be used to perform tests against hypothesis. Often we are especially interested in some moments of the sampling distribution, such as the mean and the variance. In some cases it is possible to calculate the sampling distribution from the econometric model. But sometimes, especially for finite (small) samples, this is either not possible or very difficult. In these cases Monte Carlo experiments are an intuitive way to obtain information about the sampling distribution and hence about the “quality” of the estimator.