Learn Moving Average (MA) Models

Swipe to show menu

The moving average model, often abbreviated as MA(q), is a foundational concept in time series forecasting. Unlike moving average smoothing, which is a technique used to smooth out short-term fluctuations in observed data, the MA(q) model is a statistical model that describes a time series based on past error terms. The distinction is important: moving average smoothing is a data preprocessing tool, while the MA(q) model is a probabilistic model used for forecasting.

The MA(q) model assumes that the current value of a time series can be expressed as a linear combination of the current and previous random shocks (also called error terms or white noise). The general form of an MA(q) model is:

y_t = \mu + \varepsilon_t + \theta_1 \varepsilon_{t-1} + \theta_2 \varepsilon_{t-2} + \ldots + \theta_q \varepsilon_{t-q}

where:

$y_t$ is the value of the time series at time $t$ ;
$μ$ is the mean of the series;
$ε_t$ is the white noise error term at time $t$ ;
$θ_1, θ_2, ..., θ_q$ are the parameters of the model;
$q$ is the order of the model, indicating how many lagged error terms are used.

This formulation means the present value is influenced by the current and previous $q$ shocks, rather than past observed values. In contrast, moving average smoothing simply averages observed data points over a window to reduce noise, without modeling the underlying data-generating process.

Understanding this distinction is essential for selecting the right approach for your forecasting problem. When you use an MA(q) model, you are assuming the underlying data follows a pattern where random shocks have a lasting impact for up to $q$ periods. This is fundamentally different from smoothing, which is only meant to clarify existing trends or cycles in the data.


              123456789101112131415161718192021222324252627282930
            
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from statsmodels.tsa.arima.model import ARIMA

# Simulate a time series with an MA(1) process
np.random.seed(42)
n = 200
mu = 5
theta1 = 0.7
errors = np.random.normal(0, 1, n + 1)
y = [mu + errors[0]]
for t in range(1, n):
    y.append(mu + errors[t] + theta1 * errors[t-1])
y = np.array(y)

# Fit an MA(1) model
model = ARIMA(y, order=(0, 0, 1))
fit = model.fit()

# Plot original series and residuals
fig, axs = plt.subplots(2, 1, figsize=(10, 6), sharex=True)
axs[0].plot(y, label="Simulated MA(1) Series")
axs[0].set_title("Simulated MA(1) Time Series")
axs[0].legend()
axs[1].plot(fit.resid, label="Residuals", color="orange")
axs[1].set_title("Residuals of Fitted MA(1) Model")
axs[1].legend()
plt.tight_layout()
plt.show()

Everything was clear?

Thanks for your feedback!

Section 1. Chapter 7

Ask AI

Ask anything or try one of the suggested questions to begin our chat

Section 1. Chapter 7