本研究藉由將單一交易日分割為許多區間，以收集更多高低價比值樣本，並用以計算價差估計值。研究結果顯示當區間長度超過30分鐘時，高低價差估計法的估計效率較Roll估計法為優。具體地說，當區間長度分割為50至70分鐘時，高低價差估計法的百分比誤差會降至最低。相對而言，當區間長度短於20分鐘時，Roll估計法的估計效率則優於高低價差估計法，並出現較低的百分比誤差。 By dividing a trading day into several intraday periods, a greater number of samples of high-low ratio are collected to compute a fractional high-low spread estimator. With a dataset of the length of intraday period longer than 30 minutes, fractional high-low spread estimators surpass Roll spread estimators. Specifically, to minimize percentage error of a fractional high-low spread estimator, the optimal length of intraday period should be chosen between 50 to 75 minutes. In contrast, with a dataset of the length of intraday period shorter than 20 minutes, Roll estimators outperform fractional high-low spread estimators in a lower percentage error.
