Library / Confidence Intervals Rather Than P values: Estimation Rather Than Hypothesis Testing

Reference

M J Gardner, D G Altman “Confidence intervals rather than P values: estimation rather than hypothesis testing” (1986) // BMJ. Publisher: BMJ. Vol. 292. No 6522. Pp. 746–750. DOI: 10.1136/bmj.292.6522.746

Bib

@Article{gardner1986,
  title = {Confidence intervals rather than P values: estimation rather than hypothesis testing},
  volume = {292},
  issn = {1468-5833},
  url = {http://dx.doi.org/10.1136/bmj.292.6522.746},
  doi = {10.1136/bmj.292.6522.746},
  number = {6522},
  journal = {BMJ},
  publisher = {BMJ},
  author = {Gardner, M J and Altman, D G},
  year = {1986},
  month = {mar},
  pages = {746–750}
}

Quotes (2)

Confidence Intervals

We have argued that the excessive use of hypothesis testing at the expense of more informative approaches to data interpretation is an unsatisfactory way of assessing and presenting statistical findings from medical studies. We prefer the use of confidence intervals, which present the results directly on the scale of data measurement. We have also suggested a notation for confidence intervals which is intended to force clarity of meaning. Confidence intervals, which also have a link to the outcome of hypothesis tests, should become the standard method for presenting the statistical results of major findings.

Correspondence Between Confidence Intervals and Hypothesis Tests

Although for quantitative data and means there is a direct correspondence between the confidence interval approach and a t test ofthe null hypothesis at the associated level of statistical significance, this is not exactly so for qualitative data and proportions. The reason is related to the use of different estimates of the standard error for the usual tests of the null hypothesis from those given here for constructing confidence intervals. The lack of direct correspondence is small and should not result in changes of interpretation. In addition, more accurate confidence intervals can sometimes be obtained by using estimates of the standard error of the sample statistic at the confidence limits themselves-such as derived by Cornfield for relative risks.