what's the feature of a bad weights matrix?

[udlich]

In some cases I meet two troubles:

1. \sum_n W_nk is slightly large than 1, about 1.0005
2. huge covariance

I try to inspect the weights matrix, however it's difficult to figure out the underlying defects in the weights matrix
Can you give me some advice? Thanks!

[mrshirts]

Can you try this dataset with 4.0 now? The solvers might be better able to converge. You an also try bootstrapped errors, which will be more valid when analytical errors are failing.