How should I interpret code co-occurrence?
Overview
Use Co-occurrence to examine where two codes occur together under a rule you select. Start with the rule, compare source support with repeated coding, then read the linked evidence. Find it under Analysis › Relations & Maps › Co-occurrence.
What to do
- Same document: count a source once when it contains both codes, even in different passages. Overlap: require overlapping text or time selections, codes on the same saved region, or a frame within a coded time range. Nearby: use the selected word or time distance; overlapping selections also count. Separate image regions are not compared by geometric distance.
- Qualifying sources counts each source meeting the rule once. Coding-assignment pairs counts every qualifying pair of assignments to the two codes. Several pairs can come from one source: 12 pairs in 2 sources does not mean 12 sources or 12 distinct passages.
- Worked example: 10 sources are selected. Code A occurs in 4 sources, B in 3, and both occur in 2. Thus 5 sources contain either code (4 + 3 − 2). If the Overlap rule is met in only 1 source, relationship coverage is 1/5 = 20%. Under Same document, Jaccard is 2/5 = 0.40. These answer different questions.
- By group: prevalence divides qualifying sources by all selected sources in that group, including uncoded sources. In the example above, if all 10 sources belong to one group, overlap prevalence is 1/10 = 10%, while relationship coverage remains 1/5 = 20%. A group with no qualifying pair has zero prevalence; coverage is not applicable when neither code occurs in that group.
- Read the figures: the matrix uses one low-to-high colour scale for the selected measure. 0 means no qualifying pair, … means below the display threshold, and – means an excluded comparison. The diagonal shows each code’s own source count. The network prioritises the selected measure, then qualifying-source support; display limits can leave pairs out. Inspect the pair list and evidence before choosing relationships to report.
When to use it
- To find codes that tend to appear in the same documents.
- To choose pairs worth reading in the evidence.

