Hi all,

I have a problem that seems like it should be straightforward but I'm having a terrible time searching for it.

I have a data set with known variances which is fit to a line with a weighted least squares calculation. The slope of the line from this data set is then used as a data point for further analysis, so it would be useful to have a good approximation of its variance.

However, the variance of the slope in the WLS algorithm seems to only account for the relative variances between points, and not on the overall variance. I.e., I can double the variance of each of the points and it will calculate the exact same variance for the slope of the line of best fit.

So I don't know how to propagate the overall error from the data set to the calculated slope from the WLS. What is the best way to account for this?