08 Jun 2020 (Last Modified 06 Jun 2020)
A p-value quantifies the chance that the population parameters is at least as big as the sample statistic. A p-value is often used as a proxy for how likely the results of a biomedical experiment could have risen by chance. Comparing these two sentences reveals the divide between the intended and actual uses of p-values.
P-values are misused, by those failing to appreciate the difference between quantifying the significance of an association (essentially the fraction of area under a probability density function) and quantifying the magnitude of the association. The confidence interval quantifies the precision of the estimates of the magnitude of association.
A blind application of p-values without consideration of experimental design may lead to false positives (because of failing to correct for multiple simultaneous comparisons) or failure to consider that p-values fall with increasing n, but this reflects a mathematical relationship rather than a more precise estimate of something (recall the above distinction between p-values and confidence intervals).