placebo and showed a near-significant reduction in the primary composite of total HF hospitalisations and CV death (RR 0.79, 95% CI 0.62–1.01, p = 0.059), driven by fewer total HF hospitalisations (RR 0.74, 95% CI 0.58–0.94, p = 0.013), with no difference in CV death (HR 0.96, 95% CI 0.70–1.32, p = 0.81) [ 56 ].
A field guide to the near-miss p-value
Ways to say not significant without saying not significant.
When a p-value lands just past 0.05, the result does not change but the prose does. This is a catalogue of that prose: 411 distinct phrases, found and quoted in 460,334 open-access papers, each one shown in the sentence its authors actually published.
514,635
sentences quoted
411
distinct phrasings
0.050
median reported p-value
8,873
journals
One at random
another →The wall at 0.05
Every reported p-value in the corpus, in bins of 0.01. Nothing about the statistics changes as the line is crossed. Everything about the writing does.
How the corpus lands
Most-used phrasings
all phrases →| Phrase | Sentences | Median p | Worst offender |
|---|---|---|---|
| highly significant | 132,143 | 0.001 | 1.000 |
| did not reach statistical significance | 111,028 | 0.045 | 0.050 |
| an increasing trend | 63,972 | 0.050 | 0.996 |
| showed a trend | 53,323 | 0.070 | 1.000 |
| a decreasing trend | 47,181 | 0.050 | 1.000 |
| marginally significant | 26,083 | 0.060 | 1.000 |
| a clear trend | 19,461 | 0.057 | 0.990 |
| may be significant | 13,098 | 0.050 | 0.500 |
| borderline significant | 11,410 | 0.056 | 1.000 |
| a positive trend | 11,006 | 0.061 | 0.922 |