Q2 Flashcards
Nature of unstationary data
Against critical assumption of CLRM, means variance and mean are not constant over time
Time series data should be…
Stochastic process sequence of random variables ordered in time
Test for stationary data?
Dickey Fuller or Augmented Dickey Fuller
Process of DF…
Regress first differences of the log of X on trend variable and 1 period lag
Null hypothesis of DF
B3 is zero, otherwise contains unit root / unstationary
Interpretation of DF
The T stat:
Stat < critical value ; unstationary
Stat > critical value ; stationary
Three versions of DF with equations
Random Walk: Change X = B3Xt-1 + ut
Random Walk with Drift: Change X = B1 + B3Xt-1 + Ut
Random Walk with Drift around deterministic trend: Change X = B1 + B2t + B3Xt-1 + Ut
The DF test assumes…
No autocorrelation
Can DF be used on higher than 1 level lag?
No, assumption of autocorrelation will be violated, but use ADF
What does ZA model test for?
Break point in structurally broken data on the bassi of t stat - single break point is endogenously determined
Null hypothesis for ZA test
H0: variable follows random walk with no structural breaks
Alternative hypothesis of ZA?
H1: variables are trend stationary with a one-time break and precise timing in unknown
ZA test criteria
If T stat < critical values : accept null: nonstarionary
What does Phillips-Perron test test for?
Serial correlation when testing for unit root
PP method says autocorrelation does not…
Affect asymptomatic distribution of t stat
