问题如下:
Richard Frank, FRM, is running a regression model to forecast in-sample data. He is concerned about data mining and over-fitting the data. Which of the following criteria provides the highest penalty factor based on degrees of freedom?
选项:
A.Mean squared error (MSE)
B.Unbiased mean squared error (s2)
C.Akaike information criterion (AIC)
D.Schwarz information criterion (SIC)
解释:
D is correct.
考点Selecting Forecasting Models
解析The Schwarz information criterion (SIC) has the highest penalty factor. The mean squared error (MSE) does not penalize the regression model based on the increased number of parameters, k. The penalty factors for s2, AIC, and SIC are (T/T k), e(2k/T), and T(k/T), respectively. Thus, SIC has the greatest penalty factor.
答案里给出的s2、AIC、BIC的penalty factor分别是怎么求出来的? 之前一直理解的AIC的是k前面的系数2,bic是k前面的lnT。