Talk:Breusch–Pagan test

Untitled

The page writes heteroskedasticity, but the page links to heteroscedasticity; should this be fixed?

Also, I feel tempted to add computer implementations of the test, for example, R (programming language) has a bptest function to do this test. Is this appropriate? Albmont 12:06, 10 November 2006 (UTC)

Someone else

In the introduction the test statistic is described as "n * chi^2"; I believe the "chi" in that expression should be an "R". — Preceding unsigned comment added by 216.157.201.108 (talk) 20:21, 26 February 2012 (UTC)

Need for improvement

The article doesn't specifically state what the test-statistic is, but says both that a chi-squared anf and F-test can be used. —Preceding unsigned comment added by Melcombe (talk • contribs) 13:03, 28 April 2008 (UTC)

vague writing

This article says:

Suppose that we estimate the equation

$y=x_{0}+x_{1}\beta +u.\,$

That is vague. The use of x₀ as a single term makes it look as if that's the y-intercept, so you might think x₁ is the slope and β is the independent variable. But it is customary that when lower-case Greek letters are used in this context, those are the unobservables to be estimated and the lower-case Latin letters are the observables. Whoever wrote this was confused and leaves the reader confused. Michael Hardy (talk) 19:24, 30 April 2008 (UTC)

Relation to White test?

How does this relate to the White test? 72.227.165.191 (talk) 09:59, 15 November 2009 (UTC)

Several minor problems

I agree with the above poster about the ambiguity of x₀ (wow that's easier to do in LaTeX). Also I note the following problems:

The section "Application in Economics" begins for "parameter estimation in economic time series..." Nothing about this test is specific to time series. The test applies in panel data as well as cross sectional data. If anything issues of serial correlation, not heteroskedasticity, are more relevant to time series.

Also it says generally ML or FWLS offer efficiency gains over OLS except in the presence of heteroskedasticity. Boy is that backwards! OLS is BLUE under the standard assumptions, including homoskedasticity, and there are efficiency gains under heteroskedasticity. Also Ordinary Least Squares is the Maximum Likelihood estimator under the assumption that the errors are normal and the conditional expectation is linear.

I'm going to clean this up some. — Preceding unsigned comment added by Burkander (talk • contribs) 14:00, 20 March 2011 (UTC)

Possible sources for materials to extend this article?

I just came by one (but didn't get to include the information in the test):

• http://www2.hawaii.edu/~taylor/z632/diagnosticsnotes.pdf

Please extend the list as you come along good references. Tal Galili (talk) 21:22, 22 July 2012 (UTC)