Douglas R. White. 2007. Using the Durbin-Watson test with multiple regression
Source of the Durban-Watson lookup table:
The Durbin-Watson Test for Serial Correlation with Extreme Sample Sizes or Many Regressors
N. E. Savin; Kenneth J. White Consult article at Jstor for missing values if needed http://links.jstor.org/sici?sici=0012-9682%28197711%2945%3A8%3C1989%3ATDTFSC%3E2.0.CO%3B2-D
Econometrica, Vol. 45, No. 8. (Nov., 1977), pp. 1989-1996. Charted by Number of Independent Variables in the Regression Equation
Variables= 1 Variables= 2 Variables= 3 Variables= 4 Variables= 5 Variables= 6 Variables= 7 Variables= 8 Variables= 9 Variables= 10
p<.05 dL dU dL dU dL dU dL dU dL dU
n 1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9 10 10
10 0.88 1.32 0.70 1.64 0.53 2.02 0.38 2.41 na na
20 1.20 1.41 1.10 1.54 1.00 1.68 0.79 1.99 0.69 1.99 The Durban-Watson is out of date, but it is the only option now in Spss v14.0
30 1.35 1.49 1.28 1.57 1.21 1.65 1.14 1.74 1.08 1.83 D-W tests for correlation between a residual and the immediately preceding residual.
40 1.44 1.54 1.39 1.60 1.34 1.66 1.23 1.72 1.23 1.79 The ordering of the observations should be: one-dimensional  and directed
50 1.50 1.59 1.46 1.63 1.42 1.67 1.34 1.72 1.34 1.77 (so that influence moves in only one direction along the line).
60 1.55 1.62 1.51 1.65 1.48 1.69 1.44 1.73 1.41 1.77 For a time series, both of these conditions are met. 
70 1.58 1.64 1.55 1.67 1.53 1.70 1.49 1.74 1.46 1.77 The SCCS ordering does not meetl these conditions.
80 1.61 1.66 1.59 1.69 1.56 1.72 1.53 1.74 1.51 1.77 For example, Thonga (3) is preceded by Kung Bushmen (2); but Thonga would be better paired with the culture that follows it in the list, Lozi (4).
90 1.64 1.68 1.61 1.70 1.59 1.73 1.57 1.75 1.54 1.78 Contiguity relationships really can't be collapsed to a single order dimension because
100 1.65 1.69 1.64 1.72 1.61 1.74 1.59 1.76 1.57 1.78 there are likely to be several cultures defined as contiguous to a given culture.
150 1.72 1.75 1.71 1.76 1.69 1.77 1.68 1.79 1.66 1.80 Nonetheless, your regression may have serial correlation even on this limited criterion.
200 1.76 1.78 1.75 1.79 1.74 1.80 1.73 1.81 1.72 1.82 But the test is not complete
p<.01 READING THE TABLE: For any number of Independent Variables in your Regression Equation
10 0.39 1.14 0.34 1.70 There are two values to consider for Durbin-Watson: dL and dU (lesser and upper values)
20
30 If your Durbin-Watson value is less than the dL you have serial correlation
40
50 If your Durbin-Watson value is less than the dU you probably have serial correlation
60 (especially, in this case, if one of your independent variables is a measure of time)
70
80 Durbin-Watson is used in regression analysis with a constant term and one or more independent variables,
90 where the observations are consecutive, as in the Standard Cross-Cultural Sample or time series.
100 1.52 1.56 1.48 1.60
150 1.61 1.64 1.58 1.67 Using /Analyze/Regression/Linear in Spss, the Durbin-Watson calculation is done
200 1.66 1.68 1.64 1.70 when this option is clicked in the STATISTICS box
The test results is between 0 and 4, with 2 meaning no serial correlation
In cross-cultural research, the Durbin-Watson provides Serial correlation means, for low values of, that adjacent cases are SIMILAR.
information needed for determining whether the significance However, if you subtract the values from 4, a resultant low value that is statistically significant
of a finding should be lowered by reducing the effective sample size means that adjacent cases are more DIVERSE that excected by change.
because some of the cases act as nonindependent duplicates.
This table was copied from Savin and White (1977) by DR White, 2007.
The Excel file for this Durbin-Watson.htm is on this site and named Durbin-Watson.xls
For a full treatment of this probblem see http://www.mtsu.edu/~Eeaeff/662/c09.pdf