From Wikipedia, the free encyclopedia
Jump to: navigation, search
WikiProject Statistics (Rated B-class, Low-importance)
WikiProject icon

This article is within the scope of the WikiProject Statistics, a collaborative effort to improve the coverage of statistics on Wikipedia. If you would like to participate, please visit the project page or join the discussion.

B-Class article B  This article has been rated as B-Class on the quality scale.
 Low  This article has been rated as Low-importance on the importance scale.
WikiProject Mathematics (Rated B-class, Low-importance)
WikiProject Mathematics
This article is within the scope of WikiProject Mathematics, a collaborative effort to improve the coverage of Mathematics on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
Mathematics rating:
B Class
Low Importance
 Field:  Probability and statistics



Why? Simon (reference 2) says it's common, arbitrary, and usually a mistake; Fisher (reference 4) says it's arbitrary, and convenient. (It probably wasn't common back then). I don't have a copy of Rees (reference 1).

The other references are only for notation and value. Merge to misuse of statistics? — Arthur Rubin | (talk) 00:59, 3 February 2008 (UTC)

Simon says it's "used all the time" and is "arbitrary" but nowhere says it's "usually a mistake". While his article isn't as clear as it might be, I do not see how it can be interpreted in this way. What he is arguing against is deciding post-hoc to present (say) 90% confidence intervals rather than the conventional 95% after gathering the data, and I would certainly agree.
I'm not saying it was common in 1925 when Fisher wrote his book - I'm saying the influence of this book was ultimately largely responsible for making it common.
I gave Rees as one example of an introductory textbook, partly as large parts of it are available on Google Books. A search on 1.96 "confidence interval" in Google Books gives another 542 viewable book extracts , the vast majority of which appear to be relevant. It doesn't seem to be wikipedia practice to include such links in articles as evidence of notability though. As the article was under a day old I hadn't had time to put in a whole series of references.
I'm fairly new around here, but if you wish to suggest a merge, isn't it more usual to insert a {{merge}} template and discuss it on the article's talk page rather than to {{prod}} it immediately and take it to WP:AFD within 24 hours of the article's creation? I don't understand why you suggest merging to "misuse of statistics" in particular. Are you claiming that the common statistical practice of using 95% confidence intervals is in fact a misuse of statistics? This may be a valid viewpoint, but I believe the major purpose of an encyclopedia is to describe the world as it is, rather than to try to change it. If on the other hand you're claiming it's not notable, it seems to be that this WP guideline says an attempt should be made to look for sources before nominating for deletion.
I apologise if I've done something against Wikipedia customs or practice or misinterpreted policies or guidelines. As you're a much more experienced wikipedian than me i'd be grateful if you can explain. --Qwfp (talk) 18:20, 4 February 2008 (UTC)
I originally thought it should be deleted
  1. regardless of sources; it is arbitrary and is usually mistaken, although I need to find more definite sources for that for that, and
  2. certainly should be (deleted) from Numbers, where I initially found the reference;
but now I think it should be merged to some other article, perhaps confidence interval. I'm not saying that you did anything wrong. We just have a difference of opinion about notability, and even if your google books sources are mostly more-or-less relevant, they probably shouldn't count unless the 90% and 99% confidence intervals are not present. Looking through my (advanced) textbooks, one of the mentions 1.96 as being the 95% confidence interval, but says it's arbitrary, and says nothing about common use; and that's it. The other three don't mention it at all, as far as I can tell without looking at each page. — Arthur Rubin | (talk) 18:52, 4 February 2008 (UTC)
I've just added some more refs and quotes to the article to show that 95% is a widely used, though I agree ultimately arbitrary, convention.
I can't see where it's linked from the Numbers article. I categorised it as Category:mathematical constants. I'm quite happy to remove that categorisation, as I can see now that it doesn't really fit in. I'll do that now... --Qwfp (talk) 20:20, 4 February 2008 (UTC)
Perhaps it was WT:WikiProject Mathematics. Sorry, it was your comment in Wikipedia talk:WikiProject Numbers. In any case. With your permission, I'll edit the category above so that it's a reference. I now think it should be merged to confidence intervals, I believe it's too late to withdraw the AfD. — Arthur Rubin | (talk) 20:37, 4 February 2008 (UTC)


I was trying to find the exact value of this number and this article was really helpful - Thanks! (talk) 17:45, 9 January 2009 (UTC)

Possible error in leading sentence[edit]

The article starts off as follows: "1.96 is the approximate value of the 97.5 percentile point of the normal distribution used in probability and statistics." Isn't this wrong? Shouldn't it say: "1.96 is the approximate value of the 97.5 percentile point minus the 2.5 percentile point of the normal distribution used in probability and statistics." 1.96 is meant to represent 1.96 SDs, not 1.96 SDs plus the 2.5 percentile point. Or have I misunderstood something?  — DemonicPartyHat talk 23:27, 18 May 2013 (UTC)


Is it 95% or 97.5% ? The text and picture seem to disagree. (talk) 22:21, 23 October 2014 (UTC)

This depends on if you are looking at a one-tailed or two-tailed test. 95% lies between -1.96 and 1.96, 97.5% lies between -infinity and 1.96.--Salix alba (talk): 22:31, 23 October 2014 (UTC)

SD or SEM ?[edit]

As far as I know the 95% confidence interval corresponds to +/- 1.96 * SEM and not 1.96 * SD as it is stated in the article. As I am not a statistician, I have not edited the article. Aseedb (talk) 07:43, 7 December 2015 (UTC)


I think that, in Mathematica, the correct function should be either

InverseCDF[NormalDistribution[0, 1], 0.975]


Quantile[NormalDistribution[0, 1], 0.975] (talk) 21:47, 1 July 2016 (UTC)

Fixed, thanks. Qwfp (talk) 08:42, 2 July 2016 (UTC)