2.4.10 | 95% credibility interval accuracy
We evaluated a posteriori if, in the vicinity of the two observed
datasets respectively, the lengths of the estimated 95% credibility
intervals (CI) for each parameter was accurately estimated or not (e.g.
Jay et al., 2019). To do so, we
calculated how many times the true parameter\(\left(\theta_{i}\right)\) was found inside the estimated 95% CI
[2.5%quantile\(\left(\hat{\theta_{i}}\right)\) ;
97.5%quantile\(\left(\hat{\theta_{i}}\right)\)], among the 1,000
out-of-bag NN-ABC posterior parameter estimations (Supplementary
Table S4 ). For each parameter, if less than 95% of the true parameter
values are found inside the 95% CI estimated for the observed data, we
consider the length of this credibility interval as underestimated
indicative of a non-conservative behavior of the parameter estimation.
Alternatively, if more than 95% of the true parameter-values are found
inside the estimated 95% CI, we consider its length as overestimated,
indicative of an excessively conservative behavior of parameter
estimation. For comparison, we conducted the above analysis using
instead parameters estimated under the loosing scenario Afr2P-Eur2P
(Supplementary Table S5 ).