A suitable selection of base functions for discrete wavelets transform is a difficult problem for data analysis. The grouping frequencies derived from the analysis are often dependent on the base functions and a selection of the base functions has been made mostly by trial and error. On the selection of the most suitable model from several statistical models, the model of which AIC (Akaike Information Criterion) is minimum, is selected as the suitable model. In this study, we propose the application of the AIC as the methodology for selecting the most suitable model for data compression on wavelets transform, that is, the compressions of one and two dimensional data are made by wavelets transform, and we select the minimum AIC model as the suitable one. The same method is applied to the analysis of the global wind, and we see that it can be also valid for an evaluation of the efficiency of data compression. The suitable selection of the base functions which reflects a characteristic of the data is constructed by the relation between the order of the base functions and the suitable model selected by AIC.

View full abstract