A C Osborne: However what you are actually doing is “Changing the Reality” of what those people and places actually experienced.
You are producing a “False Reality” which is destroying the very thing that Science is supposed to be about ie Describing Reality as we Know it.
Despite everyone’s best efforts, the data have random variation, where “random variation” is the variation that is not reproducible and not predictable. If you knew for a fact that a certain station never had any random variation in it at all, or that some did, or that all did, you would not adjust them. The problem is, all of them have random variation. Some of the apparent difference between two sites was caused by random processes; at the same time, some random processes, like unrecorded changes in jet streams, have produced correlations in the data of recording instruments separated by long differences. So what are you going to do: ignore the data? Deny the existence of random variation, either in all cases or some cases? Ignore the correlations caused by random events (or assert that they can’t occur or can’t be random)? And having made a bunch of decisions or claims like that, how are you going to investigate whether you have come up with accurate inferences about what is not in fact absolutely known?
If you know for a fact that a Paraguay station is perfectly accurate, you can adjust the algorithm so that it is not adjusted, but can be used in adjusting other stations known to be in error. But how do you know that? The BEST team reported the algorithm in great detail; and they reported the results of the 1/8 jackknifing procedure that shows that, over the full data set, their procedure does not produce a “false reality” that destroys the very thing they are supposed to be about. But back to a question: How can you tell that they have produced a “false reality” that is actually less accurate than any “reality” produced without a complex statistical procedure, such as accepting every station at face value? Or a procedure of willy-nilly selecting some for averaging, some for adjusting, some for treating idiosyncratically?