Talk:Durbin–Watson statistic
![]() | This article is rated C-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: | |||||||||||||||||||||||
|
??
[edit]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).
?
[edit]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)
- 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)
- 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;
- for i from 2 to T do
- 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 dpd 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)
Von Neumann's paper
[edit]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)
independence vs. no correlation
[edit]what is the null hypothesis exactly? in both the Von Neumann case and OLS case, please. Jackzhp (talk) 23:04, 11 March 2011 (UTC)