Curve fitting with variance

How can I fit my data with variance of each data point? I know someone recommend using weight setting, but how? Since a large variance might corresponds to a large uncertainty, I can't simply use variance for weight. What is much more reasonable in terms of mathematics?

Answer 1

[John D'Errico]

Why cannot you just use variance for weight? In fact, that is essentially what you DO want to do. The weight on a data point will be the inverse of the square root of the variance. So a high variance translates to a low weight on that point. This is exactly what a weight is, and how it will be interpreted.