An inlier is a data item treated as consistent with the main population under a chosen outlier criterion. The boundary is not always clear: as data becomes less dense away from a central group, points may fall between clear inliers and strong outliers. section 2.6
For numeric values, a z-score measures distance from the mean in standard deviations; a threshold determines which values remain inliers. A fixed percentile can instead misclassify inliers if it cuts through a natural gap in the data. In a boxplot, whiskers mark the inlier range under an IQR-based rule, though that rule can mislead for multimodal data. section 2.1.1 section 2.3.2
The relevant population depends on the data. For categorical values, a count threshold can separate inlier values from unusually rare ones. In well-clustered data, inliers lie near the centers of large, dense clusters; comparing a record with its own cluster may be more meaningful than comparing it with every record. section 2.2 section 3.2.3
Treat an inlier label as conditional, not as proof that a record is normal. Detector thresholds control how much borderline noise is excluded. With very small samples, genuine inliers may be nearly as rare as outliers, making both labels and scores uncertain. section 2.6 section 10.3
