What makes the kind of evidence you use
anecdotal is — the relatively small number of datasets considered — the significant
share of not fully quantitative
data — the large variability in the nature of the
data — the large biases involved in the preservation of the
data.