The tag is awful. It has 94 threads and the following excerpt:

Statistical intervals, such as confidence, credible, prediction or tolerance intervals.

but in fact we have separate well-established tags for (2k), (60), (200), and (30). I would say it makes sense to replace [interval] with more specific tags everywhere where applicable.

What is worse, is that [interval] tag is often used to refer to "interval data" -- a term that has nothing to do with any of the mentioned statistical intervals. See. e.g. 24 questions tagged with [interval] and [ordinal]. This usage is directly against the wiki excerpt. I would say that we do not need any tag for interval data because all data are interval unless stated otherwise. Alternatively, we can retag those with [interval-data]. What do you think?

| |
  • 3
    $\begingroup$ +1 There are also multiple (and hugely different) meanings to "interval data". Beginners know of "interval data" from Stevens' partial classification of measurement types (interval data are those that retain their meaningfulness under a continuous group of translations). A totally different meaning is that some datasets consist not of individual numbers, but of entire intervals of numbers representing possible values. One might also include a third concept of interval-censored data. $\endgroup$ – whuber Jan 18 '17 at 20:33
  • 1
    $\begingroup$ @Whuber In fact we have an interval-censoring tag so this third meaning is covered. It seems to me that your second meaning can be described by the same tag (or am I missing something)? But what do you think of the first meaning? Should questions such as Under what conditions should Likert scales be used as ordinal or interval data? that currently have an [interval] tag be retagged with [interval-data]? I'd say not, it's enough to have [ordinal] there. $\endgroup$ – amoeba Jan 18 '17 at 20:39
  • 1
    $\begingroup$ The second meaning is different from interval censoring: it's a (small) subject in its own right, sitting at the junction of computer science and statistics. The theory is used to prove theorems about floating-point algorithms by using intervals to represent definite bounds on the values. It has been adopted by some as an alternative to classical statistical analyses. Concerning the second issue, how about a [data-type] or [measurement-type] tag? Regardless, we would likely have to deprecate interval altogether and use its wiki excerpt to recommend alternatives. $\endgroup$ – whuber Jan 18 '17 at 20:43
  • $\begingroup$ An argument in defense of keeping [interval] in the cases where users mean [interval-censoring] is that often these questions are "My data lies in an interval; what do I do?". In such cases, if the users were familiar with the term "interval-censoring", they probably would not have this question. To be clear, I'm still for removing [interval], so I don't buy that argument. $\endgroup$ – Cliff AB Jan 22 '17 at 20:59

Thanks for the discussion. I replaced the wiki excerpt for with

PLEASE DO NOT USE THIS TAG. It's slowly being deleted. Use [confidence-interval], [credible-interval], [prediction-interval], [tolerance-interval], [interval-censoring], or [ordinal] if you want to ask about ordinal-vs-interval data.

and will probably start slowly retagging existing threads. If anybody wants to help with this, please do.

Update (Feb 12): The tag is eliminated.

| |
  • $\begingroup$ Down to 75..... $\endgroup$ – amoeba Jan 27 '17 at 22:30
  • $\begingroup$ Down to 60..... $\endgroup$ – amoeba Feb 2 '17 at 9:24
  • $\begingroup$ Down to 45..... $\endgroup$ – amoeba Feb 5 '17 at 17:17
  • $\begingroup$ Down to 30..... $\endgroup$ – amoeba Feb 8 '17 at 22:43
  • $\begingroup$ Down to 15..... $\endgroup$ – amoeba Feb 10 '17 at 9:37

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .