标题: Linear regression assumptions [打印本页] 作者: zbird2134 时间: 2011-7-11 19:51 标题: Linear regression assumptions
On P336 of V1, the bottom of page fine print says that serial correlation can also affect forecast accuracy. But, on P332 it says that serial correlation has no adverse affects on the regression coefficients. Arent these statemnts contradictory? Where does the adverse impact on forecasts come from if regression coefficients are unaffected?作者: scarecrow 时间: 2011-7-11 19:51
If the assumption that error terms are not correlated is violated, the coefficients (b0 and b1) themselves are not affected BUT each coefficient's standard error is biased and lower than should be. This causes t-statistics to be overinflated and causes type 1 error.
