Talk:Durbin–Watson statistic

Statistics Low‑importance

	This article is within the scope of WikiProject Statistics, a collaborative effort to improve the coverage of statistics 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.StatisticsWikipedia:WikiProject StatisticsTemplate:WikiProject StatisticsStatistics articles
Low	This article has been rated as Low-importance on the importance scale.

Economics B‑class Low‑importance

	Business and economics portal This article is within the scope of WikiProject Economics, a collaborative effort to improve the coverage of Economics 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.EconomicsWikipedia:WikiProject EconomicsTemplate:WikiProject EconomicsEconomics articles
B	This article has been given a rating which conflicts with the project-independent quality rating in the banner shell. Please resolve this conflict if possible.
Low	This article has been rated as Low-importance on the project's importance scale.

??

According to me, there is a misunerstanding in the attatched table of critical values (dL and dU)... lambda in this table does mean number of rows of the matrix X (including the "1" vector). Other way to say this is: lambda = number of regressors + 1 (for the model with constant term).

?

It would be very nice if someone could write pseudo code version of equation. I understand pseudo code much easier than math... 213.243.174.126 (talk) 17:57, 23 November 2007 (UTC)[reply]

When you have software package to do the work, there is no point to put code in the article. Jackzhp (talk) 22:02, 11 March 2011 (UTC)[reply]

You may have misunderstood the pseudocode concept, which is for explaining a process to humans, not to computers. Humans have no way of inferring how a process works from an installed software package, and even if you gave them the source code for a typical package it would be an order of magnitude harder to extract the central process from than pseudocode tailored for human consumption.

In this case the idea for the pseudocode is very simple: sum the squares of the T values, then sum the squares of the T − 1 differences between consecutive values, and divide the latter sum by the former to obtain the test statistic. In pseudocode,

for i from 2 to T do

num += (e[i] − e[i−1]) ** 2;

for i from 1 to T do

den += e[i] ** 2;

If before the process num and den are zero then after it num/den is the desired result. The same is true for the panel data formula when the above pseudocode is run on each panel in turn, which allows num and den to accumulate from one panel to the next, with d_pd then being the quotient of the aggregate numerator by the aggregate denominator after the process has been run on all panels.

Unfortunately those who have no trouble reading math find it hard to imagine that there are people who find pseudocode easier. Since the former seem to greatly outnumber the latter among editors of technical Wikipedia articles I wouldn't expect pseudocode to last long in any article. --Vaughan Pratt (talk) 17:54, 13 May 2012 (UTC)[reply]

Von Neumann's paper

His paper talks about the distribution of the ratio, what is its distribution? Can someone summarize it? It's a headache for me to read it. Jackzhp (talk) 23:02, 11 March 2011 (UTC)[reply]

independence vs. no correlation

what is the null hypothesis exactly? in both the Von Neumann case and OLS case, please. Jackzhp (talk) 23:04, 11 March 2011 (UTC)[reply]