Wikipedia:Articles for deletion/1.96

The following discussion is an archived debate of the proposed deletion of the article below. Please do not modify it. Subsequent comments should be made on the appropriate discussion page (such as the article's talk page or in a deletion review). No further edits should be made to this page.

The result was keep in the sense of "not delete"; no consensus yet whether it should be merged and where to. That's left as an exercise to editors. Sandstein (talk) 21:35, 8 February 2008 (UTC)

1.96

1.96 (edit | talk | history | protect | delete | links | watch | logs | views) – (View log)

Original {{PROD}} reason:Why z_.025 (5%)? Not commonly used, even in statistics. Non-notable number. — Arthur Rubin | (talk) 15:31, 2 February 2008 (UTC)

• In my work, we use z_5E-8 corresponding to a 1 - 1E-7 confidence interval. Other fields that correctly use statistics use still other confidence intervals. — Arthur Rubin | (talk) 15:33, 2 February 2008 (UTC)
  • Good for your work. The rest of the world thinks otherwise. Schaum's Outline of Theory and Problems of Probability and Statistics says that +1.96 and -1.96 are "critical values". Andy Field's Discovering Statistics Using SPSS says that 1.96 is "an important value". Michael J. Crawley in Statistics: An Introduction Using R says that these two numbers "are two very important numbers in statistics". So whom is Wikipedia to believe? Arthur Rubin who says that this number is "non-notable" and "not commonly used" simply because he personally doesn't use it in his own work, or statistics textbooks that say that it is "important" and "critical" and that use it all over the place? Wikipedia is for readers who might want to find out about the number 1.96 that they see being used all over the place, not for the Arthur Rubins of this world who don't use that number. 86.20.169.102 (talk) 20:36, 2 February 2008 (UTC)
    • Merge to misuse of statistics, perhaps? If the choice of confidence interval is not made before the experiment is performed, it should fit. — Arthur Rubin | (talk) 00:45, 3 February 2008 (UTC)
• If this number is only notable because of such interval construction, this type of discussion can be put in an article about them, or under Normal distribution, etc. but this article itself should be deleted. It isn't π, after all. Baccyak4H (Yak!) 19:24, 2 February 2008 (UTC)
• Keep but move to a more accurate name. Right out of the gate, Qwfp had three citations from professional statistician's journals. Anton Mravcek (talk) 23:47, 2 February 2008 (UTC)
  • What name do you have in mind? 95 percent confidence interval in statistics? If I could think of a name under which it would be appropriate, I would have suggested it as an alternate.
    • I don't know, that sounds better than the current name. I would not keep the article under its current name. Shalom (Hello • Peace) 01:02, 3 February 2008 (UTC)
• Merge into Normal distribution. I see little advantage to a separate article for this aspect of the subject. Tim Ross·^talk 12:21, 3 February 2008 (UTC)
• merge to Confidence region or Confidence interval 5%, 2.5%, 1%, are commonly used for confidence intervals especially in medical statistics, these become gold standards in many papers. They are convenient values as often the sample size is too small to use a smaller percentage. Merging to Normal distribution is not ideal as the 5%, 2.5%, 1% are used for a variety of statistical tests t-test, F-test etc. --Salix alba (talk) 14:25, 3 February 2008 (UTC)

p.s. Kind of curious about Rubins's work which allows such tight intervals. Care expand?--Salix alba (talk) 14:27, 3 February 2008 (UTC)

See WAAS#Integrity. I realize it's a special case. — Arthur Rubin | (talk) 15:54, 3 February 2008 (UTC)

Curiously 95% comes up in the paragraph just abouve that.--Salix alba (talk) 00:18, 4 February 2008 (UTC)

• Comment If there's a merger, confidence interval might be a better place than normal distribution, but of course the information could go into both articles. Isn't it still somewhat conventional for medical journals to treat 5% as a sort of quasi-sacred number for these purposes? Also, in introductory statistics instruction, one sees this number a lot. Michael Hardy (talk) 22:26, 3 February 2008 (UTC)
• Keep (but renaming it to 95 percent confidence interval in statistics is a good option). This number is notable, as seen from the google books results, which include some that explicitly call it "an important number". This particular degree of confidence is extremely popular in many fields, and there are probably more references about it (for example, someone must have discussed the idea that 1 out of 20 published studies is wrong just on statistical grounds! ;-) --Itub (talk) 16:55, 4 February 2008 (UTC)
• Merge (changing my deletion nomination) to confidence interval. — Arthur Rubin | (talk) 20:34, 4 February 2008 (UTC)
• Merge to confidence interval or rename to something more appropriate. -- Fropuff (talk) 06:46, 5 February 2008 (UTC)
• Redirect to normal distribution#Standard deviation and confidence intervals. This has now turned into a discussion of whether to merge, and if so to where, rather than a deletion debate. No-one seems to like the current title and I don't think there's a suitable alternative name to move it to. I admit I misread WP:NUM#Numbers in statistics as a justification for creating an article -- my apologies. If it's to be merged, I think it would be better for the redirect to go to normal distribution than confidence interval as this particular number is a property of the normal distribution, although its importance comes from the use of the normal in constructing confidence intervals. In particular, a redirect to normal distribution#Standard deviation and confidence interval seems appropriate (with a note "1.96 redirects to here"). I'd agree with Michael Hardy that some of the information could go into both articles. I don't think it's appropriate to put the whole content into either though (particularly all the notes i've put in since it was nominated for deletion to establish that 95% confidence intervals are commonly used - though one or two could go into confidence interval). I've copied the content to my user space and i'll think about what bits could usefully go into confidence interval or normal distribution. I think my time would be more usefully spent trying to improve those, so I'd be glad if this debate can be closed soon if possible. --Qwfp (talk) 16:46, 6 February 2008 (UTC) (I am the article's primary author)
• Procedural comment The nominator and I ended up having fairly civilised discussion over at Talk:1.96. I would have preferred if the debate had begun there with a {{notability}} tag on the article (in line with my reading of WP:Guide to deletion#nomination) rather than bringing the matter here within a 24 hours of the article's creation. Thanks to all above for your comments and your time. It's been a learning experience. --Qwfp (talk) 16:46, 6 February 2008 (UTC) (I remain the article's primary author)

The above discussion is preserved as an archive of the debate. Please do not modify it. Subsequent comments should be made on the appropriate discussion page (such as the article's talk page or in a deletion review). No further edits should be made to this page.