Abstract:
The Block Maxima method divides sample data into equal blocks. Predictions are based on the maximum values of the observations. Choosing an efficient and proper block size for the Block Maxima method is an important issue and varies across fields (e.g., flood, rainfall, finance). However, the main problem is deciding which block size is suitable or optimal for the prediction. In the literature, it is a known fact that the selection of a small block size leads to bias, while the selection of a large block size leads to a variance problem. In one respect, this issue is any trade off problem between the bias and the variance. This paper proposes simple and easy computational method to specify the optimal block size selection process for the Block Maxima method.
