The error involved the method in which pentad (5-day) accumulations of SSM/I emission precipitation estimates were aggregated into psuedo-calendar months prior to the original authors' (Huffman et. al., 1995) merging of these data with other sources (raingages, IR estimates, etc.). Since pentads do not generally fall on calendar month boundaries, the data must be aggregated to simulate calendar months. Unfortunately, the data from the various sources were not normalized to calendar monthly amounts in the same fashion.
The end result is that time series of global precipitation from the original data set shows erroneous maxima each August that are attributable to the way that the pentad data were aggregated into monthly amounts. This new data set remedies this problem.
REFERENCE: Huffman, G. F. (et. al.) 1995, J. Climate, vol. 8, 1284-1295.