In the initial release of FaceReader 9, an issue was identified with the Baby model: automatic analysis did not detect any negative valences across the dataset. As a result, expressions that were actually negative or very negative were misclassified as having positive valence.
No. Fortunately, re-analysis is not necessary. This is because valence can be recalculated directly from the action units, which are still accurate.
This issue has been resolved in FaceReader version 9.0.17. You can download the update from MyNoldus.