Intro to Business Statistics: Time Series Analysis Smoothing Methods Moving Averages Exponential Smoothing

What This Is

Smoothing methods, specifically moving averages and exponential smoothing, are used in business to reduce noise and variability in time series data. This helps in making informed decisions about sales forecasting, inventory management, and quality control. For instance, a retail chain wants to know if average daily sales exceed $10,000 to determine if they need to restock inventory.

Key Formulas & Symbols

• Moving Average (MA): MA = (Σx_i) / n where x_i = individual data points, n = number of data points.
• Exponential Smoothing (ES): ES = αx_t + (1-α)ES_(t-1) where x_t = current data point, ES_(t-1) = previous ES value, α = smoothing constant (0 < α < 1).
• Smoothing Constant (α): α is a value between 0 and 1 that determines the weight given to new data points.
• Weighted Moving Average (WMA): WMA = (Σw_i x_i) / Σw_i where w_i = weights assigned to each data point.
• Simple Moving Average (SMA): SMA = (Σx_i) / n where x_i = individual data points, n = number of data points.
• Exponential Smoothing Error (ESE): ESE = |ES - x_t| where ES = ES value, x_t = current data point.

Step-by-Step Procedure for Exponential Smoothing

1. Choose a value for the smoothing constant (α).
2. Set the initial value of the exponential smoothing (ES) to the first data point (x_1).
3. For each subsequent data point (x_t), calculate the new ES value using the formula: ES = αx_t + (1-α)ES_(t-1).
4. Repeat step 3 for each data point in the time series.
5. Use the final ES value to make predictions or decisions.

Step-by-Step Procedure for Moving Averages

1. Choose the number of data points (n) to include in the moving average.
2. Calculate the sum of the individual data points (Σx_i).
3. Divide the sum by the number of data points (n) to get the moving average (MA).
4. Repeat steps 2-3 for each set of n data points in the time series.
5. Use the final MA value to make predictions or decisions.

Common Mistakes

• Mistake: Using a fixed value for α without considering the trade-off between bias and variance.
• Correction: Choose α based on the specific problem and data, considering factors like data frequency and seasonality.
• Mistake: Not accounting for the initial value of ES when using exponential smoothing.
• Correction: Set the initial value of ES to the first data point (x_1) to ensure accurate calculations.
• Mistake: Misinterpreting the p-value as the probability that H₀ is true.
• Correction: The p-value is the probability of observing the data (or more extreme) if H₀ is true, not the probability that H₀ is true.

Quick Practice Problems

1. A company uses exponential smoothing to forecast sales. If the current sales data point is $100,000 and the previous ES value is $90,000, and α = 0.2, what is the new ES value? Answer: $92,000. Calculation: ES = 0.2(100,000) + 0.8(90,000) = 92,000.
2. A retail chain uses a moving average to track daily sales. If the daily sales data points for the past 7 days are $5,000, $6,000, $7,000, $8,000, $9,000, $10,000, and $11,000, what is the 7-day moving average? Answer: $7,142.86. Calculation: MA = (Σx_i) / n = (5,000 + 6,000 + 7,000 + 8,000 + 9,000 + 10,000 + 11,000) / 7 = 7,142.86.
3. A company uses exponential smoothing to forecast inventory levels. If the current inventory level is 500 units, and the previous ES value is 450 units, and α = 0.3, what is the new ES value? Answer: 475 units. Calculation: ES = 0.3(500) + 0.7(450) = 475.

Last-Minute Cram Sheet

1. Exponential Smoothing (ES): ES = αx_t + (1-α)ES_(t-1).
2. Smoothing Constant (α): 0 < α < 1.
3. Moving Average (MA): MA = (Σx_i) / n.
4. Weighted Moving Average (WMA): WMA = (Σw_i x_i) / Σw_i.
5. Simple Moving Average (SMA): SMA = (Σx_i) / n.
6. Exponential Smoothing Error (ESE): ESE = |ES - x_t|.
7. α = 0.05: Default significance level.
8. Degrees of Freedom (df): df = n - 1.
9. ⚠️ p-value is NOT the probability that H₀ is true – it’s the probability of observing the data (or more extreme) if H₀ is true.
10. ⚠️ Choose α based on the specific problem and data, considering factors like data frequency and seasonality.