Term
|
Definition
| The elasticity of Y with respect to X |
|
|
Term
| Meaning of SemiLog (lnX) B1 |
|
Definition
| The change in Y (in units) related to a 1 percent increase in X |
|
|
Term
| Meaning of SemiLog (lnY) B1 |
|
Definition
| The percent change in Y (in units) related to a 1 unit increase in X |
|
|
Term
|
Definition
| Roughly, the slope of Y with respect to X for small X |
|
|
Term
|
Definition
| Roughly, the inverse of the slope of Y with respect to X for small X |
|
|
Term
| Omitted Variable Consequences, Detection, and Correction |
|
Definition
1. Bias in the coefficient estimates
2. significant unexpected signs, poor fits
3. Include the omitted variable or a proxy |
|
|
Term
| Irrelevant Variable Consequences, Detection, and Correction |
|
Definition
1. Less precision, higher standard errors, lower t-scores.
2. Theory, t-test, R-squared, impact if dropped
3. Remove the variable |
|
|
Term
| Incorrect Functional Form Consequences, Detection, and Correction |
|
Definition
1. Biased estimates, poor fit, difficult interpretation.
2. Examine theory, think about the relationship
3. Transform to different functional form |
|
|
Term
| Multicollinearity Consequences, Detection, and Correction |
|
Definition
1. high standard errors and low t-scores
2. high correlation or high VIF
3. Drop redundant variables, increase N, or do nothing |
|
|
Term
| Serial Correlation Consequences, Detection, and Correction |
|
Definition
1. OLS no longer BLUE, hyp testing is unreliable
2. durbin-watson d test, significantly less than 2
3. Impure: change form. Pure: GLS or Newey-West SEs |
|
|
Term
| Heteroskedasticity Consequences, Detection, and Correction |
|
Definition
1. OLS no longer BLUE, hyp testing is unreliable
2. Park or White Tests
3. Impure: fix form. Pure: HC SEs or reformulate variables. |
|
|
Term
| Perfect Multicollinearity violates... |
|
Definition
|
|
Term
| A variable that explains all of the change in Y is called a... |
|
Definition
|
|
Term
| Imperfect multicollinearity is a _______ __________ as well as a _________ ___ |
|
Definition
| sample phenomenon; theoretical one |
|
|
Term
| Pure Serial correlation is a violation of... |
|
Definition
|
|
Term
| Impure serial correlation is caused by... |
|
Definition
| omitted variable or incorrect functional form |
|
|
Term
| Remedies for serial correlation |
|
Definition
GLS (Cochrane-Orcutt method)
Newey-West Standard Errors |
|
|
Term
| Pure heteroskedasticity occurs when _________ ________ _ is violated. |
|
Definition
|
|
Term
|
Definition
| the variance of the error term changes proportionally to the square of this Z |
|
|
Term
| Impure heteroskedasticity |
|
Definition
| caused by specification error |
|
|
Term
| How do we do the Park test? |
|
Definition
ln(ei^2) = a0 + a1lnZi + ui
Check significance of the t-score of a1 |
|
|
Term
| How do we do the White Test? |
|
Definition
(ei)^2 = a0 + ... all regular, squared, and cross-product variables.
TS = NR^2 with df = # of slope coefficients
#X = 1,2,3,4,5 #df = 2,5,9,14,20 respectively |
|
|
Term
| Assumptions for using durbin-watson d test |
|
Definition
1. model includes intercept term
2. et - pet-1 + ut
3. does not include a lagged dependent variable as an independent variable. |
|
|
Term
| Properties of Distributed lag models |
|
Definition
1. Severe multicollinearity
2. B0, B1, B2,... smooth declining
3. lower df, oh noooooo!!! |
|
|
Term
| Dynamic models suffer from... |
|
Definition
| serial correlation, which causes bias |
|
|
Term
| Testing for serial correlation in a Dynamic Model |
|
Definition
Lagrange Multiplier Test
Residuals = a0 + a1Xt + a2Yt-1 + a3et-1 + ut
H0: a3 = 0, Ha: a3 != 0
TS = NR^2 > chi df = # restrictions in H0 |
|
|
Term
| Correcting for Serial Correlation in Dynamic Models |
|
Definition
| Improve specification, Use instrumental Variables, modified GLS |
|
|
Term
|
Definition
Yt = B0 +B1Yt-1 + ... + BpYt-p + a1At-1 + ... apAt-p + et test At-1,...,At-p Get F statistic and compare to F critical value (K*,N-K-1) K* = # of restrictions, N = obs., K = # explanatory variables.
If we reject, then we have evidence that A granger causes Y |
|
|
Term
|
Definition
| When two variables have no causal relationship but it may be wrongly inferred that they do. |
|
|
Term
| A variable is stationary if... |
|
Definition
1. the mean of xt is constant over time
2. the variance of xt is constant over time
3. r(xt,xt-k) depends only on k for all k |
|
|
Term
| Test for non-stationarity |
|
Definition
1. |gamma| < 1 => stationary
2. |gamma| > 1 => non-stationary
3. |gamma| = 1 => non-stationary, unit root, random walk. |
|
|
Term
| Dickey-Fuller Test for non-stationarity |
|
Definition
Yt = gammaYt-1 + vt
Yt - Yt-1 = (gamma-1)Yt-1 + vt
changeYt = BYt-1 + vt, B = (gamma-1)
H0: B=0 (gamma=1,unit root)
Ha: B<0 (gamma<1,stationary)
Cannot use a t-test |
|
|
Term
|
Definition
| ChangeYt = a0 + a1ChangeXt + et |
|
|
Term
|
Definition
Yt = a0 + a1Xt + et
bot Yt, Xt unit root
et = Yt - a0 - a1Xt
Xt, Yt are cointegrated if their linear combination is stationary. |
|
|
Term
|
Definition
| dummy variable as dependent variable |
|
|
Term
| Problems with linear probability model |
|
Definition
1. y not bounded between 0 and 1
2. R^2 not an accurate measure of overall fit
Pseudo R^2 = Count R^2 = # of correct obs./#obs |
|
|
