Non Stationarity Flashcards
Types of non stationary
Deterministic
Random walk with drift
Why could regression falsely have high r squared
2 variables trend Over time
Won’t follow t distribution
Random walk model- replacing 🕤>1 with =1
as >1 has shocks and doesn’t describe many data sets
Shocks
🕤<1 desirable- die gradually
=1 random walk persist and don’t die
>1 explosive shocks
Detrending
Substitute variables- differencing- removes non stationarity removed but introduces MA(1) structure
White noice graph
⏸⏸⏸⏸⏸
Random walk graph/ with drift
⏸⏸⏸↗️↗️
Deterministic
📈
Autoregressive
⏸⏸⏸⏸⏸ for =0
↗️↘️↗️↘️↗️↘️for =1
If non stationary has more than 1 unity root
Difference more than 1 time
Non stationary has been differences d times
Order d
Formula has triangle
Test for unit root
Dickey- fuller
Formula random walk
Yt=yt-1 +ut
Formula stationary order AR(1)
Yt=🕘yt_1 +ut
Stationary AR(1) with drift
🕘yt_1 +m+ut
Stationary AR(1) with drift and time trend
🕘yt_1 +m+🔼t+ut
Augmented dickey fuller- hint fake dick
Augment using p lags
Difficult to determine optimal number of lags
Phillips Perron test
Unit root non-stationary test
Difficult to decide if 🕘=1 or 0.95
