Multiple vs simple linear regression

31 Views Asked by At

some problem which intuitively seems to work, but where I have some problems to prove. Consider two regression lines fitted by OLS, namely

$\hat{y_i}=b_0 +b_1x_{i1} + b_2x_{i2}$ (1)

and

$\tilde{y_i}=\tilde{b_0} +\tilde{b_1}x_{i1}$ (2)

where $\tilde{y_i}$ is the residual from a regression of $y_i$ on a constant and $x_{i2}$.

Show that

$\mid{\tilde{b_1}\mid}\leq\mid{{b_1}\mid}$

and how you have to change the second equation such that

$\mid{\tilde{b_1}\mid}=\mid{{b_1}\mid}$

How to approach this by using for example the standard normal equations and the resulting properties about the resdiuals? Thanks a lot!