不失一般性地, 在條件期望值函數是有興趣的條件特徵的情況下, 本文說明為什麼非實驗性資料的性質會使得隨機迴歸的誤差項有不同的意義, 從而允許了對條件期望值函數可能的錯誤設定.
Assuming without loss of generality that conditional expectation is the conditional attribute of interest, we show why the properties of nonexperimental data give rise to a different interpretation of the error term in stochastic mean regression and in turn allow possible misspecification for conditional expectation.