Jump to content

Talk:Wald test

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia


What exactly is the null hypothesis in the Wald Test? First para:"it tests whether an independent variable has a statistically significant relationship with a dependent variable."

Is it then testing the NH that X and Y are independent variables? If not, just say that. The way it is written is crap. I've changed it. —Preceding unsigned comment added by 152.16.225.228 (talk) 18:49, 24 July 2009 (UTC)[reply]

Need of Clarification

[edit]

I am not too knowledgeable on this topic, so just commenting based on a "user"s perspective...I would appreciate if it said in the section on tests on multiple parameters what the variable "n" signifies. I guess it's usually sample size, but good to be explicit in maths.152.19.16.49 (talk) 19:14, 3 September 2015 (UTC)[reply]

Need of Expansion?

[edit]

IMO I think this article would benefit if it developed the Wald test in the way it is typically used i.e. as a test of whether some linear transformation of the parameter vector is equal to the zero vector, with perhaps an example and comments on how it is frequently used as a large-sample test - that is, the version of the test that is usually written in matrix notation. I would, but I'm not sure I'm qualified enough to do a reasonable job.--Theodds (talk) 17:41, 7 December 2010 (UTC)[reply]

Name of article

[edit]

Shouldn't we move it to Wald's test for consistency with his other articles and grammatical correctness?--Adoniscik (talk) 07:01, 11 January 2008 (UTC)[reply]

It's usually referred to as a "Wald test" in statistical practice - at least in my discipline, which is medical statistics. --Pstevens (talk) 13:20, 18 April 2008 (UTC)[reply]
It's "Wald's test" in my dictionary of statistics (Everitt). Melcombe (talk) 13:37, 18 April 2008 (UTC)[reply]

Copyediting

[edit]

Thanks to the previous editor for improving the flow of the text. I've just changed the first section a little because the Wald test is not only for dichotomous variables [1]. I've mentioned continuous variables specifically. I haven't (yet) put anything about categorical variables being handled via multiple dummy variables because the page doesn't yet describe the Wald test on multiple variables. I'll come back to this when I have more time.--Pstevens 11:48, 14 November 2006 (UTC)[reply]

Stating the hypothesis explicitly

[edit]

I hope I haven't caused offence by reverting the edits by 134.48.244.232. In fact there are many hypotheses that can be tested by a Wald test, because it is a very general test procedure for parametric models. --Pstevens (talk) 13:37, 18 April 2008 (UTC)[reply]

Speaking of generality, it would be nice if there were some discussion of generalized (>1df) Wald tests.

Alternatives

[edit]

I modified the Alternatives section to remove the reference to confidence intervals. You could construct confidence intervals based on either a Wald or LRT statistic, so the confidence interval is irrelevant. — Preceding unsigned comment added by 209.6.139.32 (talk) 18:09, 23 February 2013 (UTC)[reply]

References for the Talk Page

[edit]
  1. ^ Stata 9 reference manual (Volume [R]R-Z page 416), ISBN 1-881228-93-2 (I'll look for a better reference to use in the main article.)

Asymptotic equivalence

[edit]

The article seems to imply that asymptotic equivalence of Wald, score and likelihood ratio tests was established by Robert Engle in 1983. However, this result was well-known in 1983, for instance it is covered in Chapter 9 of Theoretical Statistics (Cox and Hinkley, 1974). I believe a more appropriate primary source is; Wald, A. (1943) `Tests if statistical hypotheses concerning several parameters when the number of observations is large'. Trans Amer. Math. Soc., 54, 426-482. Aztek41 (talk) 08:58, 21 September 2023 (UTC)[reply]