When two things move together

Two quantities rising together can be the beginning of a good question. The pattern may suggest a connection worth investigating. Understanding that connection takes more than drawing a line through the points.

Describe the pattern first

Imagine a building where electricity use and visitor numbers are recorded each day. On busier days, electricity use tends to be higher. Before explaining why, look closely at the observations: are weekends different, are there unusual days, and does the pattern hold across the whole range?

A scatter plot can help reveal the shape of a relationship and observations that deserve attention. As NIST's guidance explains, it can show an association without establishing that one variable causes the other. A clear picture is a starting point for interpretation.

The description should also name what is missing. Daily totals might hide changes within each day. A short record might cover only one season. The same plotted pattern can support several explanations when the surrounding conditions are unknown.

Give the alternatives a fair hearing

Visitors might contribute to electricity use. Longer opening hours could also increase both measures. Weather might affect attendance and heating or cooling. A special event could bring more people while requiring additional lighting and equipment.

These are possible explanations for the imagined example, not findings. Listing them is useful because each suggests different information to collect. A record of opening hours would answer one question; equipment schedules or temperature records would answer others.

A third factor associated with both quantities can make a simple causal story misleading. The direction of influence may also be unclear in other settings: does one quantity change the other, does the influence run the other way, or do both respond to something else?

There is another possibility to consider before any of these stories: how the observations were selected. A pattern among recorded days need not describe days that were left out. The selection rule belongs beside the result.

Ask what would distinguish the explanations

A useful next step is to turn each explanation into a question that evidence could address. If longer opening hours explain much of the association, what would we expect when comparing days with similar hours? If equipment use matters, what additional record would make that visible?

Such comparisons still require care. Grouping observations by one condition does not guarantee that every other relevant difference has been addressed. Experimental design can strengthen causal investigation by deliberately varying conditions and planning comparisons, but the details of the design and its limitations remain important.

When reading a causal claim, it helps to ask:

Was the proposed cause observed before the effect?
What other factors could produce the same pattern?
How were observations selected and measurements made?
What comparison would make the competing explanations less plausible?

Let the wording match the evidence

“These measures varied together” and “changing this measure changed that one” make different claims. Choosing the first statement when only association has been established preserves what the observations actually show.

An association can still be useful for description, prediction, or deciding what to study next. Its value grows when we are clear about the question it answers and the questions it leaves open.