NEED TO KNOW Flashcards
How to test the null of AC?
Use the d-stat against the upper and lower bounds.
How do you run the BG test?
- Run the OLS and obtain residuals
- Regress the residuals on the regressors and the p-autoregressive terms.
- Obtain R2
What is the BG test w/ equation?
ut = put-1 + put-2 + Pput-p + vt
What is the null and alternative hypothesis for the BG test?
H0: No higher order AC - p1 = p2 = p3 = pp = 0
H1: Higher autocorrelation is present in the error terms
A consequence of MC?
Still BLUE
What is the LR relationship? Cointegration?
Each variable is non-stationarity but they have linear stationarity when combined.
What does ADF do?
Generates the first difference of the dependent variable.
What is the null of DF or ADF?
Null H0: B3=0 so unit root.
What is the difference between DF and ADF in terms of lags?
We have p-lags in ADF
What can we also do to make the data stationary?
We can take the first difference.
What is ZA?
It is a minimum root test that tests for structural breaks on the basis of t-stats
What is cointegration?
Cointegration is when one non-stationary series is regressed on another yet they do not result in spurious regression. If two-time series have a stochastic trend then this will cancel each other out (become stationary) so we have a long-run E.
How do we get the parameters in the ECM (from the cointegrating equation?)
Need to regress lc ly
How do you conduct the formal test for Cointegration?
- Determine if stationary or non-stationary?
- If non-stationary might have cointegration. - Perform cointegration equation and get ECM
- DO THIS BY REGRESSING LC LY
- Cointegrating equation is lc = b0 + b1Ly + error
- Get ECM from this error (emc) = Lc-B0-B1Ly - USE CRADF TEST TO SEE IF COINTEGRATING - THAT IS IF STATIONARY
- Identify if the dependent variable is lc or ly?
- CREATE EMC1 Lagged emc= l.emc
The function of emc1 (lagged error term) that is negative and statistically significant is the dependent.
- USE REGRESS DLC EMC1
- DLY EMC1 - Specify ARDL/ESTIMATE THE SR EQUATION
- Search for parsimonious ARDL
- This could be taking the first difference but removing the error correction model.
What do we regress to find what is dependent in cointegration?
regress dlc emc1
regress dly emc1
EMPHASIS ON THE d
In the EG test what do u need to do to actually tell if there is cointegration?
Need to use the CRADF test.
H0: Unit root - no cointegration
H1: Stationarity - we have a cointegrating relationship.
What does the ECM do? What don’t we observe?
It uses lagged error terms in the short-run to count for any disequilibrium between Yt* and Yt.
Where Yt* is the variable we don’t observe and Yt we do observe.
How is the LR achieved with SR disequilibrium?
Treat error term Ut=ln(ct) - B0 - B1(Yt) as equilibrium error term that corrects deviations of ln(Ct) from its E value.
What is the SR Granger Causality test?
Test helps us to find if LY is affected by the disequilibrium in LC.
What is the LR cointegrating relationship?
Ln(Ct) = B0 + B1ln(Yt) + Ut ln = constant + ly + error
What is regressed to determine how much the series has been corrected by?
regress dlc emc1 dlc1 dlc2 dlc3 dlc4 dly dly1 dly2 dly3 dly4
What does the coefficient of ecm1 mean?
if there was any deviation from E value in the previous period, it has been corrected in the current period by x%.
What do we run the CRADF test on?
The error term (EMC) with suppressed constant and 2 lags