Scribd
Upload
8 views

acf

Autocorrelation measures how a time series correlates with its past values at different lags, indicating relationships between current and previous time points. The autocorrelation formula calculates this relationship, with high values suggesting strong dependence, negative values indicating oscillation, and no significant correlation implying unpredictability. ACF plots visually represent these correlations, revealing patterns such as seasonality when spikes occur at regular intervals.

Uploaded by

Deepa Ravindran
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
8 views

acf

Autocorrelation measures how a time series correlates with its past values at different lags, indicating relationships between current and previous time points. The autocorrelation formula calculates this relationship, with high values suggesting strong dependence, negative values indicating oscillation, and no significant correlation implying unpredictability. ACF plots visually represent these correlations, revealing patterns such as seasonality when spikes occur at regular intervals.

Uploaded by

Deepa Ravindran
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
You are on page 1/ 2

**Autocorrelation** refers to the correlation of a time series with its own past

values, i.e., how a time series is related to itself at different time lags. In
simple terms, it's a measure of how the values at one time point relate to the
values at a previous time point.

**Lag** refers to the number of time periods between the current value and the past
value you're comparing it with. So, a "lag 1" autocorrelation means you're
comparing the time series with its value from the previous time period, "lag 2"
means you're comparing it with two time periods ago, and so on.

### Autocorrelation Formula:


Autocorrelation at lag \( k \) (denoted as \( \rho_k \)) is calculated as:
\[
\rho_k = \frac{\sum_{t=1}^{N-k} (X_t - \bar{X})(X_{t+k} - \bar{X})}{\sum_{t=1}^{N}
(X_t - \bar{X})^2}
\]
Where:
- \( X_t \) is the value at time \( t \).
- \( \bar{X} \) is the mean of the time series.
- \( k \) is the lag (i.e., the difference in time between the two values you're
comparing).

### 📊 What Does Autocorrelation Tell Us?


- **High Autocorrelation at Lag 1**: If there is strong autocorrelation at lag 1,
it suggests that today's value is similar to yesterday's value. In financial data,
for example, today's stock price might be highly correlated with yesterday's price.

- **Negative Autocorrelation**: If the autocorrelation is negative, it means that


when one value is high, the next value tends to be low, and vice versa. This could
indicate a cyclical pattern or oscillation in the data.

- **No Autocorrelation**: If there's no significant autocorrelation at a given lag,


it means there's no predictable pattern between the current value and the lagged
value.

### 🌟 Example of Autocorrelation in Time Series:


Let's consider a **stock price time series**:

- **Lag 1 (Autocorrelation at lag 1)**: How does today’s stock price relate to
yesterday's stock price?
- **Lag 2 (Autocorrelation at lag 2)**: How does today’s stock price relate to two
days ago?
- **Lag 3 (Autocorrelation at lag 3)**: How does today’s stock price relate to
three days ago?

If you plot the **autocorrelation function (ACF)** for different lags, you'll see
how each lag impacts the correlation. A spike at **lag 1** indicates that the stock
prices are similar from day to day.

---

### 📈 ACF Plot Interpretation:


An **ACF plot** shows the autocorrelation at each lag.
- The **x-axis** represents different lags (time intervals between observations).
- The **y-axis** represents the correlation values for each lag.

#### Example:
1. **Stock price data**: If the ACF shows a high positive correlation at lag 1, it
suggests that the stock price today is highly dependent on the stock price
yesterday.
2. **Temperature data**: If you’re looking at daily temperature data and see
significant autocorrelation at lag 365, this could indicate yearly seasonality, as
temperatures tend to follow a similar pattern each year.

---

If there are **spikes at regular intervals** in an autocorrelation plot (ACF plot),


it typically indicates **seasonality** or a repeating pattern in the data at those
intervals.

### In simple words:


- A **spike** means that the data at a certain time is highly correlated with data
at another specific time.
- If these spikes occur at **regular intervals**, it suggests that the data follows
a predictable pattern that repeats itself over time. For example, this could be
yearly, monthly, or weekly patterns.

### Examples:
- **Daily Sales Data**: If you see spikes at **every 7th day** (lag 7), it may
indicate weekly seasonality, like high sales on weekends.
- **Monthly Temperature Data**: If there are spikes at **lag 12**, it could
indicate yearly seasonality, such as temperature being similar month-to-month every
year.
- **Stock Price Data**: Spikes at regular intervals may show patterns like daily
price trends or monthly patterns.

### What does it mean?


- If you see these regular spikes in the ACF plot, it’s a clear indication that the
data has **seasonal behavior** or repeats itself after a fixed period.
- These spikes tell you that the past values influence future values after a
consistent interval, and this interval is the **seasonal period**.

---

You might also like