📌 Chapter 2: Fundamental Concepts

1. Autocorrelation Function (ACF)

γk
ρk =
​

γ0
​ ​

ρk = Autocorrelation at lag k (how related a value is to its past values).

​

γk = Covariance at lag k (measures how two values change together).

​

γ0 = Variance (how spread out the data is).

​

2. Partial Autocorrelation Function (PACF)

No fixed formula but calculated by removing intermediate lag effects.
Helps find direct relationships between a point and its past values.

3. White Noise Process

et ∼ iid(0, σ 2 )
​

et = Error term (random noise).

​

iid = Independent and Identically Distributed (each value is unrelated to others).

σ 2 = Variance (spread of the noise).

4. Random Walk Process

Yt = Yt−1 + et
​ ​ ​

Yt = Value at time t.
​

Yt−1 = Previous value.

​

et = Random error (new unpredictable change).

​

Used in stock prices and natural movements.

5. Moving Average (MA) Process

Yt = et + θet−1
​ ​ ​

Yt = Value at time t.
​

et = Current error term.

​

θ = Weight of previous error term.

Used to smooth fluctuations in data.

6. Autoregressive (AR) Process

Yt = ϕYt−1 + et
​ ​ ​

Yt = Value at time t.
​

Yt−1 = Previous value.

​

ϕ = Coefficient (strength of relationship between past and present).

et = Random error.
​

7. First Differencing for Stationarity

ΔYt = Yt − Yt−1
​ ​ ​

ΔYt = Change between current and past value.

​

Used to remove trends and make data stationary.

8. Augmented Dickey-Fuller (ADF) Test

p
ΔYt = α + βt + γYt−1 + ∑ δi ΔYt−i + et
​ ​ ​ ​ ​ ​

i=1

α = Constant.
βt = Time trend (if present).
 γYt−1 = Checks if the time series is stationary.
​

δi ΔYt−i = Past differences.

​ ​

et = Random error.
​

If γ is significantly negative, the series is stationary.

9. KPSS Test (for stationarity)

Tests if a series is stationary or not.
If p-value < 0.05, the series is non-stationary.

📌 Chapter 3: Exponential Smoothing

1. Simple Exponential Smoothing (SES)

St = αXt + (1 − α)St−1
​ ​ ​

St = Smoothed value at time t.

​

Xt = Actual observed value.

​

α = Smoothing parameter (how much weight recent data gets).

St−1 = Previous smoothed value.
​

2. Double Exponential Smoothing (Holt’s Method)

St = αXt + (1 − α)(St−1 + bt−1 )

​ ​ ​ ​

bt = β(St − St−1 ) + (1 − β)bt−1

​ ​ ​ ​

St = Smoothed value.
​

Xt = Actual value.
​

α = Level smoothing factor.

bt = Trend estimate.
​

β = Trend smoothing factor.

3. Triple Exponential Smoothing (Holt-Winters Method)

Additive Model (for constant seasonal effects)

St = α(Xt − Ct−m ) + (1 − α)(St−1 + bt−1 )

​ ​ ​ ​ ​

bt = β(St − St−1 ) + (1 − β)bt−1

​ ​ ​ ​

Ct = γ(Xt − St ) + (1 − γ)Ct−m
​ ​ ​ ​

St = Smoothed value.
​

Xt = Observed value.
​

Ct = Seasonal component.
​

α = Level smoothing factor.

β = Trend smoothing factor.
γ = Seasonality smoothing factor.
m = Number of periods in a season.

Multiplicative Model (for increasing seasonal effects)

Xt
St = α ( ) + (1 − α)(St−1 + bt−1 )
​

Ct−m
​ ​ ​ ​

bt = β(St − St−1 ) + (1 − β)bt−1

​ ​ ​ ​

Xt
Ct = γ ( ) + (1 − γ)Ct−m
​

St
​ ​ ​

Same as above, except seasonality is proportional to trend.

Summary Table
Formula Meaning

St = αXt + (1 − α)St−1
​ ​ ​

Simple Exponential Smoothing (SES)

St = αXt + (1 − α)(St−1 + bt−1 )

​ ​ ​ ​
Holt’s Method (Double Exponential Smoothing)

St = α(Xt − Ct−m ) + (1 − α)(St−1 + bt−1 )

​ ​ ​ ​ ​
Holt-Winters (Additive Seasonality)

St = α ( CXt−mt ) + (1 − α)(St−1 + bt−1 )

​
​

​ ​ ​
Holt-Winters (Multiplicative Seasonality)
​
Formula Meaning

Yt = Yt−1 + et
​ ​ ​
Random Walk (non-stationary process)

Yt = ϕYt−1 + et
​ ​ ​

Autoregressive Process (AR)

Yt = et + θet−1
​ ​ ​

Moving Average Process (MA)

ΔYt = Yt − Yt−1
​ ​

First Difference (for stationarity)