Not infallible!
A possible merge was labeled "These 2 people have a high degree of inconsistency. Merging is not recommended." However, the only inconsistency was less than a year difference in the birth year, and that was because one of the birth year dates was a census estimate. Most of the time the warnings are good, and make me review the data. I also like having to give a reason when I decide to merge despite warnings. But this particular dire warning was uncalled for. 😊
Comentários
-
You should probably list the two IDs of the profiles that were merged so that the programmers can analyze the situation and see if the warning was appropriate or not.
Any type of routine that tries to provide warnings like those in the new merge routine is going to have to obey basic statistical laws. One of these is the interplay between sensitivity and specificity. This is the relationship between true positives, false positives, false negatives, and true negatives which is that as the number of false positives go down, the number of false negatives go up and as the number of false negatives go down, the number of false positives go up. Stated another way, you cannot have both a high rate of true positives and a high rate of true negatives. It's a balancing act and the trick is finding the right balance.
1