1. Keep every intended tree observation
Preserve the names, order and complete Newick text of the selected trees. The supplied-tree path in the dated source reviewed here requires at least two parsed trees, at least four taxa and matching taxon sets. The retained public case uses exactly these three named entries:
| Name | Newick | Saved split |
|---|---|---|
| First | ((A,B),(C,D)); | AB|CD |
| Repeat | ((A,B),(C,D)); | AB|CD |
| Third | ((A,C),(B,D)); | AC|BD |
First and Repeat are separate observations with identical text. Keep the repetition if it belongs to the intended collection. Removing it changes the collection and its denominator. Conversely, an accidental duplicate should be resolved before interpretation; the program cannot establish whether a repeated tree was scientifically intended.
The source accepts this named collection through the tree_ data contract. The retained check supplied that contract directly. It did not exercise a file chooser, launcher or hosted upload. Confirm the available input path before treating the example as a GUI recipe.
2. Select the frequency question explicitly
The displayed route label is Consensus Network. All twenty routes initially start selected in the reviewed source; the saved check selected only this route. Its other nineteen route-call counts are zero. Keep the selected-method list with the input collection.
This route uses the split sets of all supplied tree entries, including repeats. A split is counted at most once within each tree’s set. This differs from pairwise Robinson–Foulds distance, which compares the first two supplied trees. The branch lengths are not weights in this split-membership frequency.
3. Read the numerator and denominator together
For each unique split, the route counts how many selected trees contain it and divides by the number of selected trees. Here the denominator is three observations, not two distinct topologies. The saved frequencies and the earlier independent reference agree on the following values:
| Split | Retained reference | Saved numeric value |
|---|---|---|
AB|CD | 2 / 3 | 0.6666666666666666 |
AC|BD | 1 / 3 | 0.3333333333333333 |
These values are dimensionless proportions of the supplied observations. The example contains no bootstrap generation, sampling model or calibration that would turn them into confidence or posterior probabilities. An observed frequency alone does not establish a tree’s biological correctness.
Match a frequency by its split identity, not by its row position: the implementation builds the union from sets, whose iteration order is not a stable reporting contract. Keep both sides of the split label with the number.
4. Preserve the complete five-field summary
The saved accumulated table contains all five fields below. The numerical checks cover split count, frequency and threshold count. Density is retained as an observed string; adjacency, density correctness and drawing were outside the accepted check.
| Saved field | Saved value | Interpretation boundary |
|---|---|---|
| Total Splits | 2 | Two distinct splits occur in the selected collection. |
| Splits > 50% | 1 | One saved split frequency is strictly greater than one half. |
| Density | 0.250 | Observed text only; the network adjacency and density were not independently validated. |
| Mean Freq | 0.500 | Three-decimal string for the mean over the unique split frequencies, not confidence in a consensus tree. |
| Ref | Holland & Moulton, J Math Biol 2003 | Recorded citation text, preserved rather than silently corrected. |
The condition behind Splits > 50% is strictly greater than 0.5; a value exactly equal to one half does not enter that count. The saved log also reports zero frequencies strictly greater than 0.75. These counts describe the saved frequencies. They do not establish that the displayed adjacency was filtered to those thresholds.
The Mean Freq field has already been formatted to three decimal places. Preserve the full split values separately when available. This saved summary has no per-split rows, original-tree text or input provenance embedded in its five fields.
5. Keep fallback controls separate from input workflows
Two retained direct calls checked internal two-set fallback shapes. They bypassed the public input-loading path. They help explain a boundary in the implementation, but they are not two additional native or alignment workflows:
| Direct control | Saved frequencies | Splits > 50% | Limit |
|---|---|---|---|
| Legacy pair with the same split | AB|CD = 1.0 | 1 | One direct route call; no public loader. |
| Alignment-shaped pair with missing per-tree split sets | AB|CD = 0.5AC|BD = 0.5 | 0 | The pair’s split sets were supplied; alignment parsing, distances and tree inference were not exercised. |
The fallback selects the two pair split sets when the collection is absent or any per-tree split set is None. The second control supplies that internal shape directly. It does not verify an end-to-end NJ or UPGMA inference path.
6. Retain evidence without inventing an export
The reviewed run saved an observation receipt containing the complete input, split labels, full numeric frequencies, summary tables, logs and call counts. It performed zero native exports and no graphics. The evidence download below makes those retained values reviewable; it is not a CSV produced by the app’s Export All action.
For a later authorized workflow, retain the complete input collection, selected route, source identity and all actual output files. Verify an export against its full numerical source before relying on it. This guide provides no evidence that this route’s native export contains per-split values or original trees. Other Tree routes’ export receipts do not fill that gap.