Linear regression with measurement errors, am I doing it right?

This looks like this discussion. I would suggest adding a parameter that scales the value_se rather than using it as-is.