DW is generally used to test the first-order autocorrelation.
LM can be used to test first-order and higher-order autocorrelation. Generally in eviews, we only test the second order. We can check the DW statistics in the regression results with lag period of 2. If it is very close to 2, it means that the autocorrelation is eliminated. At the same time, check the chi-square percentile distribution table to see if LM=(T*R square) is greater than the critical value. If so, it means that there is second-order autocorrelation.
Then the generalized difference method is used to eliminate autocorrelation.