The “weighting” that I was referring to applies to the application of the regression procedure when applied to a single sequence of temperatures (either anomalies or to the raw series as described in the post). To account for the differences in monthly variation, we can estimate trends using a weighted regression using a weight for each month that is proportional to the inverse of the variance of the series of monthly averages for that month. It is not related to the situation of averaging several series together as you refer to. Weighted averaging of series would be used if one were to calculate *area weighted* averages however the weights in such a case would be determined by the areas represented by each sequence and not by the properties of the series themselves.

As far as averaging series with missing values, I don’t find any of the methods involving anomalies particularly appealing. I use my own method which allows for missing values, does not require a common overlap period for all of the series and results in an average temperature series for which anomalies can be calculated should they be desired.

]]>In this case, I had asked someone else to look at it so I had to post it publicly with password protection.

]]>