Time series Revision Notes

A time series graph is a type of line graph used to display data points collected or recorded at regular time intervals, such as days, months, or years. The primary purpose is to show how a variable changes over time, making it easier to detect trends, seasonal effects, or irregular fluctuations.

Unlike bar charts or scatter plots, time series graphs emphasise the chronological order of data, which is essential for understanding temporal patterns. The horizontal axis (x-axis) represents time, progressing from left to right, while the vertical axis (y-axis) shows the variable being measured, such as temperature, sales, or population.

• Time series graphs display data in chronological order.
• They help identify trends, cycles, and seasonal variations.
• The x-axis always represents time intervals.
• The y-axis represents the measured variable.

Before plotting, gather the data points with their corresponding time intervals. Ensure the time intervals are consistent (e.g., daily, monthly, yearly) and clearly labelled. Choose an appropriate scale for both axes to accommodate the range of data values and time periods.

For example, if you have monthly sales figures for one year, label the x-axis with months from January to December and the y-axis with sales amounts, scaled to cover the minimum and maximum sales values. Accurate scaling ensures the graph is easy to read and interpret.

• Collect data points with consistent time intervals.
• Label the x-axis with time periods in order.
• Label the y-axis with the variable and appropriate scale.
• Ensure scales cover the full range of data values.

Plot each data point by locating its position according to the time on the x-axis and the corresponding value on the y-axis. Use a clear mark, such as a dot or small cross, to represent each point.

For example, if sales in March were 150 units, find March on the x-axis and 150 on the y-axis, then mark the point where these coordinates meet. Repeat this for all data points to represent the entire dataset.

• Locate each time interval on the x-axis.
• Find the corresponding data value on the y-axis.
• Mark the intersection point clearly.
• Plot all points before connecting them.

After plotting all points, connect them in chronological order using straight lines. This line helps visualise how the variable changes over time, highlighting increases, decreases, or stable periods.

For instance, if sales rose steadily from January to June and then declined, the connected line will slope upwards initially and then downwards, making the trend easy to observe.

• Connect points in order of time using straight lines.
• Lines illustrate the direction and rate of change.
• Helps identify upward, downward, or stable trends.
• Avoid connecting points out of chronological order.

Once the graph is complete, analyse it to identify patterns such as trends (long-term increase or decrease), seasonal effects (regular fluctuations at specific times), or irregular variations (unexpected changes).

For example, a graph showing monthly ice cream sales might reveal higher sales in summer months, indicating a seasonal pattern. Recognising these patterns aids in understanding the data and making predictions.

• Look for overall trends in the data.
• Identify any repeating seasonal patterns.
• Note any sudden or irregular changes.
• Use the graph to support data-driven conclusions.

A moving average is a statistical technique used to analyse time series data by calculating the average of a fixed number of consecutive data points. This average then 'moves' forward through the data set, one point at a time, producing a series of averages. For example, a 3-point moving average takes the first three data points, calculates their average, then moves one point forward to average the next three points, and so on.

This method smooths out short-term fluctuations or random noise, making it easier to observe the underlying pattern or trend in the data.

• Calculated by averaging a fixed number of consecutive data points.
• The 'window' size (number of points averaged) is chosen based on the data and purpose.
• Produces a new series of averaged values that smooth the original data.

Time series data often contain irregular ups and downs caused by random events or measurement errors. These short-term variations can obscure the true behaviour of the data over time. Moving averages reduce these fluctuations, highlighting the longer-term trend or pattern.

For example, monthly sales figures might vary due to promotions or holidays. Calculating a moving average smooths these effects, helping businesses understand whether sales are generally increasing, decreasing, or stable.

• Smooth out random short-term fluctuations.
• Reveal underlying trends or cycles.
• Improve clarity for interpretation and forecasting.
• Support better decision-making by focusing on significant patterns.

To calculate a moving average, decide on the number of consecutive data points to include (called the 'window size'). Then, for each position in the data series where the window fits, calculate the average of those points.

For instance, consider the data points: 12, 15, 14, 18, 20, 22, 19. To find the 3-point moving averages:

Average of first three points: (12 + 15 + 14) ÷ 3 = 13.67 (rounded to two decimals) Next three points: (15 + 14 + 18) ÷ 3 = 15.67 Next three points: (14 + 18 + 20) ÷ 3 = 17.33 Next three points: (18 + 20 + 22) ÷ 3 = 20.00 Last three points: (20 + 22 + 19) ÷ 3 = 20.33 This produces the moving average series: 13.67, 15.67, 17.33, 20.00, 20.33, which is smoother than the original data.

• Choose window size (e.g., 3 points).
• Calculate average for each consecutive group.
• Produce a new series of averages.
• Use consistent rounding for clarity.

Moving averages help identify whether data values are generally rising, falling, or stable over time. They can also reveal cyclical patterns or seasonal effects when combined with other techniques.

However, moving averages lag behind actual data because they are based on past values. This means sudden changes or turning points in the data may appear delayed in the moving average series. Understanding this limitation is important when using moving averages for forecasting or decision-making.

• Highlight long-term trends by smoothing data.
• Reduce noise but introduce lag in detecting changes.
• Useful for identifying patterns not obvious in raw data.
• Should be combined with other analysis methods for best results.

Moving averages are widely used in economics, finance, environmental science, and other fields to analyse time series data. For example, stock market analysts use moving averages to identify trends in share prices, while meteorologists use them to smooth temperature data.

When applying moving averages, the choice of window size affects the smoothness and responsiveness of the result:

Smaller windows respond more quickly to changes but smooth less. Larger windows smooth more but respond slower to changes. Selecting an appropriate window size depends on the data characteristics and analysis goals.

• Common in financial and environmental data analysis.
• Window size balances smoothness and responsiveness.
• Helps in forecasting and trend identification.
• Requires understanding of data context for effective use.

A moving average is a technique used to smooth time series data by calculating the average of a fixed number of consecutive data points. This reduces the impact of short-term fluctuations and highlights longer-term trends or cycles in the data.

For example, a 3-point moving average takes the average of every set of three consecutive values in the series. This process creates a new series of averaged values that are less erratic than the original data.

• Moving average order: the number of consecutive points averaged (e.g., 3-point, 4-point).
• Purpose: to smooth data and reveal trends by reducing noise.

To calculate a moving average of order n, sum n consecutive data points and divide by n. Then move one point forward and repeat until the end of the series.

For example, given monthly sales data: 120, 130, 125, 140, 135, a 3-point moving average is calculated as follows:

First average: (120 + 130 + 125) ÷ 3 = 125 sales units, aligned with the 2nd data point (130). Second average: (130 + 125 + 140) ÷ 3 = 131.67, aligned with the 3rd data point (125). Third average: (125 + 140 + 135) ÷ 3 = 133.33, aligned with the 4th data point (140).

• Odd-order moving averages align with the middle data point of the group.
• Moving averages reduce short-term fluctuations, making trends clearer.

When the moving average order is even, the average lies between two data points rather than directly on one. This misalignment can cause difficulties when comparing the moving average to the original data.

Centreing corrects this by averaging pairs of consecutive moving averages to shift the values to the correct position in the time series.

For example, with a 4-point moving average, the first average is between points 2 and 3. To centre, average the first and second 4-point moving averages to place the centred value at point 3.

• Centreing is only necessary for even-order moving averages.
• It involves averaging pairs of moving averages to align with original data points.
• This adjustment ensures meaningful comparison and interpretation.

Consider the data: 10, 14, 13, 15, 16, 18, 20. Calculate the 4-point moving averages:

MA1 = (10 + 14 + 13 + 15) ÷ 4 = 13, between points 2 and 3.

MA2 = (14 + 13 + 15 + 16) ÷ 4 = 14.5, between points 3 and 4. MA3 = (13 + 15 + 16 + 18) ÷ 4 = 15.5, between points 4 and 5. MA4 = (15 + 16 + 18 + 20) ÷ 4 = 17.25, between points 5 and 6. To centre these moving averages: Centre1 = (13 + 14.5) ÷ 2 = 13.75, aligned with point 3. Centre2 = (14.5 + 15.5) ÷ 2 = 15, aligned with point 4. Centre3 = (15.5 + 17.25) ÷ 2 = 16.375, aligned with point 5. This centred series can now be compared directly with the original data points at positions 3, 4, and 5.

• Calculate moving averages over even number of points.
• Average consecutive moving averages to centre.
• Align centred averages with original data points for analysis.

Moving averages smooth the data, making it easier to identify trends by reducing random fluctuations.

Centreing ensures that the smoothed values correspond correctly to the time points in the original data, which is essential for accurate trend analysis and forecasting.

Without centreing, even-order moving averages can misrepresent the timing of trends, leading to incorrect conclusions.

• Use moving averages to identify underlying trends.
• Apply centreing to even-order moving averages for correct alignment.
• Interpret centred moving averages in the context of the original time series.

Time series data records values at successive points in time, such as monthly sales or daily temperatures. Such data often contains short-term fluctuations caused by random variation or seasonal effects, which can obscure the overall direction or trend.

A trend represents the long-term movement in the data, showing whether values generally increase, decrease, or remain steady over time. Identifying this trend is crucial for understanding patterns and making forecasts.

• Trends show the general direction of data over time.
• Short-term fluctuations can hide the true trend.
• Smoothing techniques help reveal the underlying trend.

A moving average is a method to smooth data by averaging a fixed number of consecutive data points. This reduces the impact of random fluctuations and highlights the underlying trend.

For example, a 3-point moving average takes the average of data points 1 to 3, then moves one point forward to average points 2 to 4, then 3 to 5, and so on. Each average is plotted at the centre of the points used, if centring is applied, or at the last point if not.

This process creates a new series of smoothed values that can be graphed to reveal the trend more clearly.

• Moving averages smooth data by averaging consecutive points.
• Common window sizes are 3 or 5 points.
• Each moving average value represents a smoothed data point.
• Centring moving averages places the average at the midpoint of the data used.

To calculate a moving average, decide on the number of points to include (the window size). For a 3-point moving average:

1. Add the first three data values together and divide by 3.

2. Move one data point forward and repeat for the next three values, continuing until the end of the data set is reached (where possible).

• Example: Given data points: 12, 15, 14, 18, 20
• 3-point moving averages:
• (12 + 15 + 14) ÷ 3 = 13.67
• (15 + 14 + 18) ÷ 3 = 15.67
• (14 + 18 + 20) ÷ 3 = 17.33
• Plot these averages to smooth the data.

Once moving averages are calculated, plot these values on the same time scale as the original data. Connect the points with a smooth line to form the trend line.

This trend line shows the general direction of the data:

An upward slope indicates an increasing trend over time, such as rising sales or temperatures. A downward slope indicates a decreasing trend. A roughly horizontal line suggests a stable or no clear trend. Interpreting the trend line helps in understanding long-term behaviour and supports forecasting.

• Plot moving average values on the graph.
• Connect points smoothly to form the trend line.
• Use the slope to identify increasing, decreasing, or stable trends.
• Trend lines reduce noise and clarify data direction.

Moving averages smooth data but can also delay the detection of changes because they average past values. The choice of window size affects smoothness and responsiveness: larger windows smooth more but respond slower to changes.

When data has strong seasonal patterns, moving averages alone may not fully reveal trends without further analysis.

Always label graphs clearly and ensure consistent time intervals for accurate interpretation.

• Larger moving average windows increase smoothness but reduce responsiveness.
• Moving averages may lag behind sudden changes.
• Seasonal effects may require additional analysis.
• Clear labelling and consistent intervals are essential.

Seasonal variation occurs when data shows a consistent pattern that repeats at regular intervals within a year, such as monthly or quarterly changes. For example, retail sales often increase in December due to holiday shopping and decrease in other months. This pattern is predictable and differs from random fluctuations or long-term trends.

Seasonal variation is important because it affects how we interpret data over time. Recognising these patterns helps separate regular seasonal effects from other changes in the data, such as trends or irregular events.

• Seasonal variation repeats within a fixed period, usually a year.
• It causes predictable rises or falls in data at specific times.
• It differs from trends, which are long-term changes.
• Ignoring seasonal variation can lead to incorrect conclusions.

Data tables often present values recorded at regular intervals, such as monthly sales figures over several years. By comparing values for the same month across years, we can spot seasonal patterns. For example, if sales are consistently higher in July each year, this indicates a seasonal effect.

Time series graphs plot data points against time, making seasonal patterns visually apparent. Peaks and troughs occurring at regular intervals suggest seasonal variation. For instance, a graph of electricity usage might show higher consumption every winter month.

These tools help distinguish seasonal variation from random noise or trends.

• Compare same-period values across years in data tables.
• Look for repeating peaks and troughs in time series graphs.
• Use graphs to visualise the timing and magnitude of seasonal effects.
• Check for consistency of patterns over multiple years.

Moving averages smooth out short-term fluctuations to reveal underlying patterns like trends and seasonality. A moving average is calculated by averaging a fixed number of consecutive data points and sliding this window along the data series.

To isolate seasonal components, first calculate a centred moving average to estimate the trend-cycle component. Then, divide the original data by the moving average to find the seasonal effect for each period.

For example, if monthly sales in January are 120 units and the centred moving average for January is 100 units, the seasonal component for January is 120 ÷ 100 = 1.2, indicating sales are 20% above the trend in January.

• Calculate centred moving averages to estimate trend-cycle.
• Divide original data by moving averages to find seasonal indices.
• Average seasonal indices for each period over multiple years.
• Express seasonal components as ratios or percentages.
• Use seasonal components to adjust data for seasonal effects.

Seasonal components quantify how much each period's data deviates from the overall trend due to seasonal effects. A seasonal component greater than 1 (or 100%) means the data is above the trend for that period; less than 1 means below.

For example, a seasonal component of 0.85 for February indicates values are typically 15% below the trend in that month. Understanding these helps explain why data rises or falls at certain times and supports better decision-making.

Seasonal components are dimensionless ratios, so they can be compared across different datasets or units.

• Seasonal components show relative increase or decrease from trend.
• Values above 1 indicate above-trend periods; below 1 indicate below-trend.
• They help explain regular fluctuations in data.
• Useful for adjusting data to remove seasonal effects.

Once seasonal components are calculated, they can improve forecasts by adjusting trend estimates to reflect expected seasonal effects. For example, if the trend forecast for July sales is 200 units and the seasonal component for July is 1.3, the adjusted forecast is 200 × 1.3 = 260 units.

This method assumes seasonal patterns remain stable over time. It is important to check for changes in seasonal behaviour before applying past components to future forecasts.

Accurate seasonal adjustment leads to better resource planning and decision-making in businesses affected by seasonal demand.

• Multiply trend forecasts by seasonal components to get seasonal forecasts.
• Assume seasonal patterns are consistent over time.
• Check for changes in seasonal patterns regularly.
• Improves accuracy of short-term forecasts.

A trend line summarises the general direction of data points over time, showing whether values tend to increase, decrease, or remain stable. It smooths out short-term fluctuations to reveal the underlying pattern. For example, if monthly sales figures steadily rise over several years, the trend line will slope upwards. Drawing a trend line can be done by eye or by using moving averages to reduce noise.

A trend line is essential for forecasting because it provides a baseline expectation of how the data behaves over time, ignoring seasonal or irregular variations.

• Trend lines show long-term direction in data.
• They help separate overall movement from short-term changes.
• Moving averages can assist in drawing accurate trend lines.

Seasonal components represent regular, repeating patterns within specific time intervals, such as months or quarters. For instance, ice cream sales might peak every summer and dip in winter. Seasonal effects cause predictable fluctuations around the trend line.

To calculate seasonal components, data is grouped by season (e.g., all Januaries, all Februaries) and averaged after removing the trend effect. This isolates the seasonal influence, which can then be added or subtracted when forecasting future values.

Recognising seasonal components is crucial because ignoring them can lead to inaccurate predictions, especially when data shows strong periodic variation.

• Seasonal components capture repeating periodic fluctuations.
• They are calculated by averaging detrended data for each season.
• Seasonal adjustments improve forecast accuracy.

Forecasting future values involves extending the trend line beyond the observed data and adjusting for seasonal components corresponding to the forecast period. The general approach is:

1. Determine the trend value at the future time point by extending the trend line.

2. Identify the seasonal component for that time period (e.g., the seasonal index for July). 3. Combine these by adding (or multiplying, depending on the model) the seasonal component to the trend value to get the forecasted value. For example, if the trend predicts 100 units sold in July and the seasonal component for July is +20 units, the forecast is 120 units sold in July.

• Forecast = Trend value + Seasonal component (additive model).
• Forecasting requires extending the trend line beyond known data.
• Seasonal adjustments tailor forecasts to expected periodic changes.

This forecasting method assumes that the underlying trend and seasonal patterns observed in historical data will continue unchanged into the future. It presumes no structural breaks, such as economic shifts, policy changes, or unusual events that could alter the data behaviour.

It also assumes that the seasonal pattern is consistent in magnitude and timing each cycle. For example, the peak sales in summer are expected to occur every year with similar intensity.

These assumptions are necessary for the mathematical combination of trend and seasonal components to produce meaningful forecasts.

• Past trends and seasonal patterns will continue unchanged.
• No unexpected events will disrupt the data pattern.
• Seasonal effects are consistent in timing and size.

Forecasts are inherently uncertain because real-world conditions can change. Unexpected events like economic crises, natural disasters, or changes in consumer behaviour can invalidate assumptions, causing forecasts to be inaccurate.

Seasonal patterns may evolve over time, becoming stronger, weaker, or shifting in timing, which the model may not capture if based on past data alone.

Trend lines may oversimplify complex data, missing nonlinear patterns or sudden changes. Therefore, forecasts should be interpreted with caution and updated regularly as new data becomes available.

• Unexpected changes can reduce forecast accuracy.
• Seasonal patterns may change over time.
• Trend lines may not capture all data complexities.
• Regular updates and validation improve forecast reliability.