More variants of linear regression needed

We need more variants of linear regeressions:
1. Regression for values with known normal errors. Here we'll obtain not only fit result but estimate of errors and χ² goodness of fit test. I think we need multivarite normal distributuion and special data type for description of values with normal error.
2. Regression with linear combination of several functions. (In R it's known as `glm`). It's another way to look at same regression problem but I think it deserves addition to the library as well.