Identification Method - Climate Change: Identification and Projections - page 74

Furthermore, since we only have a limited number of reconstructions
(four for temperature and four for irradiance), the number is insufficient for
the law of large numbers to compensate this type of error, on which we have
no a priori statistical characteristic (variance, independence). There would
therefore be no sense in making averages from the estimates made, or in
attempting to extract likely values or ranges of uncertainty. Indeed, alleged
probability assessments would be worth no more than the scale of experts'
judgments, which are bereft of any statistical basis.
In such a situation, one solution would be to present all the results of
identification, each one being accompanied by a region of confidence and a
comment (implicit or explicit): “ on the proviso that the original data is
correct”; i.e. without any noticeable errors of calibration or alignment.
Everyone then remains free to use the results as they see fit and to draw their
own conclusions from them.