Autoregressive Models

russian salsa, classical cumbia, havana electropop

Listen on 93

Lyrics

[Verse 1]
Stock prices whisper secrets from yesterday
Each tick depends on what came before the break of day
Like echoes bouncing through financial halls
Today's return builds on history's calls
AR one connects the dots in sequence
Past performance leaves its blueprint

[Chorus]
Auto-regressive, looking backward
Current equals past plus random factor
Phi coefficient shows the power
Yesterday's grip on this very hour
Memory fading, coefficients tell the tale
How much persistence will prevail

[Verse 2]
Take today's price minus yesterday's close
Divide by yesterday, that's how the formula goes
This return depends on prior returns we've seen
Plus epsilon white noise, keeping data clean
Phi between zero and one means stable ground
Above one, explosive growth will be found

[Chorus]
Auto-regressive, looking backward
Current equals past plus random factor
Phi coefficient shows the power
Yesterday's grip on this very hour
Memory fading, coefficients tell the tale
How much persistence will prevail

[Bridge]
Estimate phi through ordinary least squares
Regress today on yesterday, see what it declares
High phi means momentum carries strong
Low phi means shocks don't linger long
Unit root test checks if phi equals one
Stationarity's where good models begun

[Verse 3]
AR two adds another lag in the mix
Two periods back help current period fix
Multiple coefficients sum must stay below
One point zero or the model will explode and grow
Volatility clustering shows in squares
AR-GARCH captures what standard models miss upstairs

[Final Chorus]
Auto-regressive, looking backward
Current equals past plus random factor
Sum of phi's must stay inside the boundary
Keep the time series foundationary
Memory patterns tell tomorrow's story
Built on yesterday's financial glory

[Outro]
When current values chase their history
AR models solve the mystery

← Time-Series Regression and Trends | Random Walks and Unit Roots →