Statistics
Lemma frequency and dispersion
Normalized, reproducible measurements for a shareable selection of works.
Inflected forms are grouped under their dictionary headword: the one Sôtêr, the automatic lemmatiser, proposes first (the first Sôtêr candidate). Words it could not lemmatise still count in the total, but are left out of the list.
Corpus selection
No work selected. Choose works in the catalogue first.
Results
Lemma statistics
| Lemma | Occurrences | Per million | Documents | DP norm. | Juilland D | Log ratio | FDR q |
|---|
Definitions and interpretation, including the abbreviated columns
- DP norm.
- Deviation of proportions, normalized: 0 means the item is spread evenly across the selected works, 1 that it sits in a single one.
- Juilland D
- A second dispersion measure, read the other way round: 1 means evenly spread, 0 concentrated in one work.
- Log ratio
- The target work against the rest of the selection, as a log2 relative frequency. Empty when no target work is chosen.
- FDR q
- Benjamini–Hochberg false-discovery rate: the share of the items flagged at this level that would be flagged by chance alone.
Per-million frequencies use all Greek tokens in the selected editions. DP compares the observed distribution with document sizes and is normalized to its attainable maximum; lower values indicate broader dispersion. Juilland’s D is computed from document-relative frequencies; higher values indicate broader dispersion.
The optional target comparison reports a smoothed log2 relative frequency and the 2 × 2 log-likelihood statistic. The displayed q value applies the Benjamini–Hochberg correction across every tested item. These tests describe this selection only and do not establish authorship or influence.