I am facing the following statistical problem:

**Background:**

I have a typical situation of multivariate linear regression analysis: several x-variables, one y-variable. The used numerical recipes code does its job perfectly. So far, so good.

**Problem:**

The situation is not this simple. My basic data are measurement data, with known errors in both the x-variables and in the resulting y-variable. The errors are dependent on the respective variables.

I have been reading and googling for a while and have found some information about 'errors-in-variables-models'. As it seems to be a current mathematical R&D topic, I have found it hard to catch the state of the art.

**What I'm looking for:**

- information about best errors-in-variables-methods for my multivariate linear approach?
- Do you know about an appropriate coded algorithm / function like numerical recipes code?

cool.bambus