📊 Data
📅Time Series Graphs
Sort the data by time and connect it into a line — rises, seasonal rhythms, and cherry-picked windows all become visible.
Temperature, stock prices, a shop's monthly sales — whenever a number changes over time, the most natural picture is to lay it out along the calendar and connect it into a line. The line tells a story: when it climbed, when it fell, and when it turned around on schedule. Bar graphs and scatter plots are good at other jobs; "how it changes over time" belongs to the time series graph.
Why the points get connected
A time series graph has time on the horizontal axis and the value on the vertical, with the monthly points joined in order. One ice cream shop's year looked roughly like this:
| Month | Jan | Apr | Jul | Oct | Dec |
|---|---|---|---|---|---|
| Sales (bars) | 200 | 800 | 2400 | 900 | 300 |
Reading a small graph
Connect the five points: from January to July the line climbs steadily, peaking at 2400 bars in July, then slides back to 900 by October and lands at 300 in December. The verdict is not "bad business" — it is more sales in summer, fewer in winter, with the pattern repeating every year.
The connecting line is not decoration: the steepness of a segment is the speed of change, and the steeper it is, the faster the climb.
Reading three kinds of movement
- Rising trend: the whole line steps upward;
- Falling trend: the whole line drifts down, with small rebounds that do not change the story;
- Seasonal swings: ups and downs cycle on a fixed schedule — the ice cream peak returns every summer.
Read the big direction first (draw a mental straight line from the first point to the last), then look for cyclical wiggles, and do not let one odd month steer you.
Can you trust the middle of the line?
The segment between two neighboring points says "we believe the change is roughly continuous": from 800 in April to 2400 in July, May and June most likely climb somewhere in between. That inference is called interpolation, and it is usually reasonable. Two things the line does not vouch for: extending it past the last point (extrapolation), and connecting across a gap where data is missing. The farther you extrapolate, the closer you get to fortune-telling.
Who chose the time window
The sneakiest trap in a time series is the cherry-picked window: plot only June to July and "sales tripled"; plot only July to December and "sales collapsed 87%". Both sentences are true, and both are lies. When you see a dramatic rise or fall, ask first — which month does this line start from? More tricks wait for you in the lesson on misleading graphs.
y = x + 2
y-intercept
(0, 2)
x-intercept
(-2, 0)
Real data zigzags, so people often summarize its direction with one straight line, called a trend line. Drag and : is the slope of the trend — positive means a slow climb, negative a gentle slide, and 0 is a line lying flat.
Check yourself
Quick quiz
1. What lives on the horizontal axis of a time series graph?
2. Extending the line past the last point is called what, and can you trust it?
3. Announcing 'sales tripled' using only June to July — what is wrong?