Data Analytics: Foundations to Practice · Correlation, Relationships, and Causation
Leading and Lagging Relationships Over Time
Even a genuine, real relationship between two variables can be misread if the time lag between them isn't accounted for, since comparing the wrong time periods against each other can hide a real relationship or fabricate a false one.
Some genuine relationships between two variables unfold with a meaningful time delay rather than occurring simultaneously; marketing spend in one month, for instance, might genuinely drive sales that materialize primarily one or two months later rather than in the exact same month the spend actually occurred, since customers often take real time to research, decide, and eventually complete an actual purchase after first being influenced by a specific piece of marketing.
- Some real relationships unfold with a time delay, like marketing spend driving sales that materialize one or two months later, not simultaneously.
- A leading indicator changes before the outcome it predicts; a lagging indicator changes only after the outcome has already occurred.
- Comparing two variables at the same time period when a real lag exists can produce a misleadingly weak or entirely absent apparent correlation.
- Testing correlation at different specific time lags can reveal the actual point at which two variables show their strongest genuine relationship.