6
$\begingroup$

I just noticed that there is a tag - it does not have a description and it is hard to guess what is its purpose? Is it about confidence intervals (it has multiple questions tagged both [confidence] and ). If yes, then it should be marked as synonym. However if it is not so, then it seems to be ambiguous and is seems to be used incorrectly.

Should we do anything with this tag?

| |
$\endgroup$
  • 2
    $\begingroup$ It certainly sounds as it is potentially very ambiguous. However, doing "something" with this tag would probably require somebody to go through all of the questions (below 100, so doable), checking the actual usage and retagging questions as appropriate. Perhaps retagging to [confidence-interval] is not necessary, as these two tags can be declared synonyms, but I am afraid that there is a fair amount of [confidence] questions that are not about [confidence-interval]. I have not checked though. Perhaps you can take a look and report on the actual usage? $\endgroup$ – amoeba Jul 13 '15 at 16:49
  • 2
    $\begingroup$ Would you say that you... lack confidence in the tag? $\endgroup$ – shadowtalker Jul 21 '15 at 23:07
  • $\begingroup$ Hi Tim. I am trying to see if we (as a community) can manage to push some tag synonyms without mod intervention. Could you see if you can upvote some of these suggestions: stats.stackexchange.com/tags/… $\endgroup$ – amoeba Jan 20 '17 at 12:13
  • $\begingroup$ up-vote in what sense? I can see them, that's all. $\endgroup$ – Tim Jan 20 '17 at 12:28
  • $\begingroup$ The interface is very un-intuitive. For each suggestion, you should click on the master tag (left column); a new page will open where you can see synonym suggestion(s), each with up/down voting arrows. There you can vote. If you don't have >=5 rep points in the master tag then you will get a pop-up saying that you cannot vote. Otherwise your vote will count. Thanks! $\endgroup$ – amoeba Jan 20 '17 at 14:07
  • 1
    $\begingroup$ @amoeba I didn't know that, done! $\endgroup$ – Tim Jan 20 '17 at 14:42
4
$\begingroup$

There are currently 123 threads tagged with .

In the next several days I plan to look through all 30-40 that are not explicitly about CIs via search

[confidence] -[confidence-interval] -[interval] -"confidence interval"
-"confidence intervals" -"confidence level" -"95%" is:question

I will retag those that are not about confidence intervals. For the ones that are about "confidence" in an output of a classifier I am/will be using the tag, as suggested in the @AdamO's answer.


UPDATE: the above search yields zero results. All the remaining [confidence] threads are about CIs. Hence, we can now make a synonym $\to$ and merge it straight away.

Please vote for this synonym here: https://stats.stackexchange.com/tags/confidence-interval/synonyms if you have >=2.5k rep and >=5 rep in the [confidence-interval] tag.

UPDATE2 (Jan 26): Synonym implemented via voting without mod intervention.

| |
$\endgroup$
7
$\begingroup$

We have 94 questions tagged , in which 19 are tagged [confidence] + [confidence-interval].

Regarding 75 questions tagged [confidence] but not [confidence-interval], at least 41 has to do with confidence intervals.

So far, we have 60 questions out of 94 tagged with [confidence] and which relates to confidence intervals - 64% from total.

Now, if we exclude [confidence] questions which were closed or which contain the words 'band', 'limit' (confidence bands; confidence limits), there are 32 posts to investigate more closely.

| |
$\endgroup$
  • $\begingroup$ +1. We should really get rid of this tag. By now there are 124 questions, most of which (about 100) are about confidence intervals. On many others the tag can be removed altogether or replaced by [probability]. There are very few questions like this one stats.stackexchange.com/questions/45174 that talk about people's confidence in their choices, and here I am not sure what to do. Perhaps replacing with [uncertainty] as AdamO suggested is fine. But [confidence] should be eliminated, it's too easy to confuse with confidence-interval and people do it often. $\endgroup$ – amoeba Jan 18 '17 at 13:55
  • 1
    $\begingroup$ Perhaps we can clean [confidence] questions of everything that is not about CIs and then make it a synonym with merge? This should be relatively easy. $\endgroup$ – amoeba Jan 18 '17 at 13:59
  • 1
    $\begingroup$ You have a lot on your plate with tag work. I will try to look into it. $\endgroup$ – amoeba Jan 18 '17 at 14:02
  • $\begingroup$ I had forgotten about this answer of mine. Thanks for your help. @amoeba $\endgroup$ – Andre Silva Jan 18 '17 at 15:02
  • 1
    $\begingroup$ See update to my answer. Please vote for this synonym and for other pending ones too (see my conversation with Tim in the comments under his question). $\endgroup$ – amoeba Jan 20 '17 at 16:54
5
$\begingroup$

It's interesting. I find a post like this one Confidence measures for Gaussian mixture models to be an appropriate use of the word confidence, in the sense that they are interested in reporting the uncertainty around a number of predicted clusters using a mixture model, but an interval is not an appropriate reporting tool.

I think the vast majority of these posts can be classified under either "confidence interval" or "hypothesis testing". I think posts such as the above could be classified as uncertainty. Conversely, perhaps no tag is necessary in that case. Any statistical practice should have an associated method to measure uncertainty... that's basically what we do as statisticians.

| |
$\endgroup$

You must log in to answer this question.

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