36-226 HOMEWORK 8 Solutions
Due: Friday 4/13/12 at 9:30am in class
Remember to show all work; just writing down the answer will not receive credit.
1. We’ve defined the simple linear regression model as:
Yi = β0 + β1
...
36-226 HOMEWORK 8 Solutions
Due: Friday 4/13/12 at 9:30am in class
Remember to show all work; just writing down the answer will not receive credit.
1. We’ve defined the simple linear regression model as:
Yi = β0 + β1Xi + i
where E[ i] = 0, Var[ i] = σ2 where i; j are uncorrelated.
(a) Find E[Yi] and Var[Yi]. Noting that E[Yi] changes with i, find E[Y¯ ] and Var[Y¯ ].
E[Yi] = E[β0 + β1Xi + i]
= E[β0 + β1Xi] + E[ i]
= β0 + β1Xi + 0
= β0 + β1Xi
Var[Yi] = Var[β0 + β1Xi + i]
= Var[β0 + β1Xi] + Var[ i]
= 0 + σ2
= σ
2
E[Y¯ ] = E "n1 Xi=1 n Yi#
= β0 + β1X¯
Var[Y¯ ] = Var"n1 Xi=1 n Yi#
Rebecca Nugent, Department of Statistics, Carnegie Mellon University, 2012 1
Note that in the calculation of the variance of Y¯, we are able to write
VarPn i=1 Yi = Pn i=1 Var[Yi] because for any i 6= j, Yi is uncorrelated with Yj.
This is a result of the fact that for any i 6= j, i is uncorrelated with j.
(b) We also showed that the least squares estimates of β0; β1 are:
β^0 = Y¯ - β^1X¯ β^1 =
PXiYi - nX¯Y¯
PXi2 - nX¯2
One property of the least squares estimates is that they’re unbiased.
Show that E[β^1] = β1 and E[β^0] = β0.
E[β^1] = E PPn i=1 n i=1XXiYi2i--nnXX¯¯2Y¯
= β1
Rebecca Nugent, Department of Statistics, Carnegie Mellon University, 2012 2
This study source was downloaded by 100000784424693 from CourseHero.com on 05-02-2021 12:49:14 GMT -05:00
https://www.coursehero.com/file/8965534/HW8Sol/
E[β^0] = E[Y¯ - β^1X¯]
= E[Y¯ ] - E[β^1X¯]
= β0 + β1X¯ - XE ¯ [β^1]
= β0 + β1X¯ - β1X¯
= β0
(c) We estimate the E[Yi] for a value Xi by the predicted value Y^i = β^0 + β^1Xi.
Then the residual is ei = Yi - Y^i.
Show that the sum of the residuals is zero, i.e. Pn i=1 ei = Pn i=1(Yi - Y^i) = 0
nX i
=1
ei =
[Show More]