دانلود رایگان مقاله لاتین داده پایدار با پیش بینی مونت کارلو از سایت الزویر
عنوان فارسی مقاله:
ارزیابی پیش بینی مونت کارلو با داده های پایدار
عنوان انگلیسی مقاله:
Monte Carlo forecast evaluation with persistent data
سال انتشار : 2017
برای دانلود رایگان مقاله داده پایدار با پیش بینی مونت کارلو اینجا کلیک نمایید.
بخشی از مقاله انگلیسی:
5. Inverting the Meese-Rogoff puzzle
Simply put, the Meese-Rogoff puzzle is the fact that a random walk model provides better forecasts of real exchange rates (ϵ) than a model based on fundamentals. Thus, a structural forecasting model of the quarterly real Deutsche Mark–US Dollar exchange rate from 1973 to 1998 is defined as ϵt+h = β0 + ρϵt + β1Xt + νt+h, (22) where X is the real interest rate differential between the two countries. The benchmark forecasting model is the random walk for forecasting the real exchange rates: ϵt+h = ϵt + νt+h. (23) Because of scale-invariance, MC random walk draws with a unit variance are appropriate. The MC test is based on one-step-ahead predictions at the 5% level, and all parameters for the alternative model are estimated via OLS. The number of in-sample observations is a rolling window of 60 quarters (R = 60 and P = 42). Our Monte Carlo test method fails to reject the hypothesis of a random walk, with p-values of 0.2125 and 0.225 based on the MSEt and ENCt statistics, respectively. Our Monte Carlo test method extends naturally to a Monte Carlo inversion framework for constructing an exact confidence set for the benchmark model parameters. Our MC inversion method sets the parameters ρ, β0, and β1, to known values ρ¯, β¯ 0, and β¯ 1, respectively, so that the null and alternative hypotheses are redefined as H0 : ρ = ¯ρ and β0 = β¯ 0 and β1 = β¯ 1 HA : ρ ̸= ¯ρ or β0 ̸= β¯ 0 or β1 ̸= β¯ 1. (24) The random walk model is clearly a special case, where ρ = 1, β0 = 0 and β1 = 0.The MC test method is applied to the benchmark model or null defined by Eq. (24). The forecast evaluation statistic from the data is obtained by imposing the null for the benchmark model, and the alternative model is estimated. The Monte Carlo series and statistic are constructed by imposing the null model. Screening over a reasonable range of parameter values, the confidence set is constructed by retaining all points that are not rejected under the null at the desired significance level.
برای دانلود رایگان مقاله داده پایدار با پیش بینی مونت کارلو اینجا کلیک نمایید.
کلمات کلیدی:
