Power Transformations in Time-Series Models of Quarterly Earnings per Share.
针对许多公司季度每股收益时间序列，数据表明在模型中引入幂变换是合适的。实证结果显示，这通常能提高预测质量，使预测结果比未变换的模型更接近财务分析师的预测。
Abstract ABSTRACT: For many quarterly time series of corporate earnings per share, the data indicate the desirability of incorporating a power transformation into the time series model. Our empirical results suggest that, for such series, this will generally lead to forecasts of improved quality. The resulting forecasts compare more favorably with those of financial analysts than do forecasts derived from models without the transformation parameter.