{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreie6o3cnv3z5ny6zxsyknp62u3dwax6auwwm3kouucllsiuz4g34ny",
    "uri": "at://did:plc:wwyqal4cnqhuwyacdj7rqq3n/app.bsky.feed.post/3mfhrzg5pq3v2"
  },
  "path": "/t/change-the-range-not-the-language-on-confidence-intervals/27740#post_16",
  "publishedAt": "2026-02-22T17:52:31.000Z",
  "site": "https://discourse.datamethods.org",
  "tags": [
    "CI"
  ],
  "textContent": "karlamoPA:\n\n> it would be useful to students of science (and scientist-public communications in general) to provide a link to a document that explains common statistical terms in plain language. Such as for CI\n\nThe link above gets the CI vaguely defined and lacks conceptual clarity: “ _A confidence interval provides the range of values, calculated from the sample, in which we have confidence that the true population parameter lies_ “ This is not true of the realized interval and as the author states, once the interval is realized, it can only be right or wrong and therefore confidence has nothing to do with this realized interval.\n\nThis interval is actually a range of **test hypotheses** under each of which the study test statistic falls in the central 95% of all test statistics specified under those test hypotheses. In simple terms, the interval denotes the range of test hypotheses under which the study data will not be unusual. By “ _not unusual_ ” it means that the study effect falls within the central 95% of values in the distribution of each test hypothesis. Note that the mean of the test hypothesis is what is listed in the interval and defines several estimates for the population parameter. This interval is therefore a range of uncertainty in our estimate of the population parameter and should rightly be called the uncertainty interval, because we are not confident about any single value in the range. There is a parallel between the significance test and the uncertainty interval in that the interval contains exactly those specific mean values of the hypothesized sampling distribution representing hypothesized parameter values that, when treated as test hypotheses, fail to be rejected by a two-sided significance test at the 5% level. It is the inversion of a family of significance tests and translates a decision threshold (“reject or fail to reject” say at 0.05 threshold) into a range of hypothesized parameter values that remain plausible, given the data, at the chosen significance level. For example, if a 95% interval was 2 to 6 then any test hypothesis between 2 and 6 will return a fail-to-reject P-value of ˃ (1 – 0.95). This means that the uncertainty interval is much more informative than the test of significance and should be prioritized in research reporting.",
  "title": "Change the range not the language on confidence intervals"
}