Talk:Normal probability plot

Q-Q Plot

A normal probability plot is a type of Q-Q plot, at least that's what the probability plot page says. That isn't mentioned anywhere in this page.

Fixed. --WBTtheFROG (talk) 20:48, 30 July 2012 (UTC)

Interpretation

It would be nice to add a section guiding interpretation...e. g. "if your plot looks like this, this is what it means." The examples cover that to a certain degree, but it could be more well organized and serve as a better guide for how to interpret various shapes. Somebody with more advanced experience than I should write that, or links should be added to existing interpretation guides if they're out there.

--WBTtheFROG (talk) 20:51, 30 July 2012 (UTC)

P(Z<z[i])?

The formula for $P(Z<z_{i})$ given in the article 2014-08-01 was the value $z_{i}$ such that:

$P(Z<z_{i})={\begin{cases}1-0.5^{1/n}&{\text{for }}i=1\\[8pt]0.5^{1/n}&{\text{for }}i=n\\[8pt]{\frac {i-0.3175}{n+0.365}}&{\text{otherwise}}\end{cases}}$

This formula appears with no derivation or citation in section "1.3.3.21. Normal Probability Plot" of the NIST / SEMATECH Engineering Statistics Handbook ("1.3.3.21. Normal Probability Plot", NIST/SEMATECH e-Handbook of Statistical Methods, NIST / SEMATECH (published 2003-06-01), 2013-10-30, retrieved 2014-08-01). [That section cites Chambers (1983), which appears to reference Chambers et al. (1983, ch. 6), which gives U[i] = (i-0.5)/n, not this.]

Moreover, I don't know a citation for this formula, and and it conflicts with the formula used by the "qqnorm" and "ppoints" functions in the stats package in R (programming language). R is increasingly the platform of choice of people worldwide engaged in new statistical algorithm development. The core developers of R are all leading experts in statistics and are typically quite careful in citing sources. In particular the documentation for "qqnorm" cites "ppoints", which in turn cites 2 other references: Becker, R. A., Chambers, J. M. and Wilks, A. R. (1988) The New S Language. Wadsworth & Brooks/Cole and Blom, G. (1958) Statistical Estimates and Transformed Beta Variables. Wiley; see [1]. This uses

(i-a)/(n+1-2*a) with a = 3/8 if n <= 10 and a = 0.5 otherwise.

Becker, Chambers and Wilks (1988) gives this formula, and cite Blom (1958); I have not checked Blom.

The core R team go out of their way to conform to international standard where they exist and to document what they do in any case.

The present article article should cite a source, especially if it uses something different from R. If I don't see a reply to this soon, I plan to revise this article to present the formula used by R, as just described.

However, the differences are slight and would be difficult to detect in normal usage. DavidMCEddy (talk) 21:20, 31 July 2014 (UTC) // revised 19:59, 1 August 2014 (UTC) and DavidMCEddy (talk) 02:57, 2 August 2014 (UTC)