We came across a rather long-toothed retraction in the Journal of the American Statistical Association, which represents a case of doing the right thing (similar to that involving the apparent first-ever English language retraction from 1756, about which we wrote in 2012).
The 1927 notice came in the form of a letter by C. H. Whelden Jr., who was for a time the chief statistician for the American National Red Cross, referencing his 1926 article in the JASA,”The Trend-Seasonal Normal in Time Series:”
My present position on the question of the trend-seasonal normal in time series is not that expressed in my article in this JOURNAL for September, 1926. I have come to realize that the criterion there presented is not general as it should be but specific in respect of arbitrary periods of time, being based upon a careless misuse of the phrase “average activity for the year.”
I must apologize for going hastily into print prior to sufficiently careful reflection upon the matter printed. With the concept of “average activity for the year,” I hope, properly in mind, I withdraw my objection to the usual method of combining measures of trend and seasonal.