Factors estimated from large sets of economic data can sometimes be less useful for forecasting, even when more data is used. This can happen when errors in the data are related to each other, or when a factor that works well in a small dataset doesn't perform as well in a larger one. In a study, using just 40 carefully chosen data series can give better forecasting results than using all 147 series. Also, adjusting the data based on their characteristics can improve forecasts. The study looked at how these factors are calculated and found that considering certain types of errors and giving more weight to certain data groups can lead to better results.