MIDTERM:
Unbiasness
– if expected value of estimator = population characteristic than its unbiased
Consistency
– if the sample size becomes large, the distribution of the estimator collapses to a spike located at the true value of the dist.
Efficient
–
estimator has minimum variance
R
2
–
Greater Variation in Q but not P?
Greater variation = R
2
should fall since TSS is larger compared to ESS
Assumptions:
1)
X causes Y, Y doesn’t Cause X
2)
Correctly measured
3)
NonRedundant, Variation in X:
ERROR Ui:
T/F
1)
If there’s no variation in Dep. V. you can’t estimate slope?
FALSE, slope=0
2)If there’s no variation in Ind. V you can’t estimate slope?
TRUE, undefined 0/x
3)The VAR of error term differs across all observations in OLS called independence?
FALSE, called homoskedasticity or VAR of error term is constant
4)Supply f(x) is perf. Elastic (flat) & varies overtime. You can estimate model with P as depend. V. and Q as indp. V.?
FALSE, Q is explained by P, not visaversa.
TRUE, E(X+Y)=E(X)+E(Y)=10
Short Answer:
1)
Assume that you estimated GPA=3 + .05*Hours. Your get BMW at least a 3.5.
You hate study and use equation to study 10hrs/week. You get a 4.3. U think maybe you
studied too much. You remember:
1)This is just a forecast
2)Error Term, actual GPA will vary upon variation of error term
3)coeffecient estimates differ for each peep
2)Suppose R.V. has PDFof f(x)=a for x’s b/w 10 & 6 & =0 otherwise.
This is the end of the preview.
Sign up
to
access the rest of the document.
 Spring '08
 Staff
 Regression Analysis, Variance, TSS, error term

Click to edit the document details