KwCheng's blog

The Stochastic error term又稱為noise component , 

在regression 中 , 我們會想要藉由多個變因來去估計出可以的結果 , 

像 Y = B0*X1 + B2*X2 , 

(假使Y是Salary , 那B1可能是年資 , B2可能是階級)

但是我們所取的變因可能無法涵蓋所有可能的因素 , 所以估計出來的Y會與實際上的Y有所誤差 , 

因此 , 在regression equation上我們通常會加上一個stochastic error term來代表這樣的差距 , 

也就是:

Y = B0*X1 + B2*X2 + e 

Stochastic error term具有幾個特性:

The stochastic error term is a random variable. Its value cannot be a priori determined.

• May (probably) contains variables not explicit in model
• Even if all variables included, will still be some randomness
• Error may also include measurement error
• Ockham’s razor: a model is a simplification of reality. We don’t necessarily want to include every explanatory variable

ref : http://wiki.answers.com/Q/What_is_stochastic_error_term

ref : http://www.bionicturtle.com/how-to/article/regression-stochastic-error-term-and-squares-quantitative/