On this page
Inspect the standardized ratio
A difference of 0.01 divided by standard error 0.002 gives statistic 5. A difference of 1 divided by standard error 0.2 also gives statistic 5. Under the same normal test, both produce the same tail probability, despite the hundredfold difference in magnitude.
Retain the effect’s original units
| Difference | SE | Standardized statistic |
|---|---|---|
| 0.01 units | 0.002 units | 5 |
| 1 unit | 0.2 units | 5 |
Ask a separate practical question
Whether 0.01 units matters depends on the task. A research report might compare the estimate and interval with a prechosen economically meaningful magnitude. A trading application also needs costs, execution and out-of-sample behavior; a small p-value alone supplies none of those inputs.
Avoid the opposite inference
A noisy study can fail to reject even when its point estimate is large. That does not establish equivalence or a negligible effect. Report uncertainty and the design’s limitations rather than reducing both significant and nonsignificant results to a usefulness label.