1. Supply the trees and support values separately
The retained cases use the same two named six-taxon trees, First and Second. Both have the complete Newick text:
((A,B),C,(D,(E,F)));
The saved parser identifies three shared splits: ABCD|EF, ABC|DEF and AB|CDEF. The support values arrive separately through nj_ and upgma_. With supplied trees, the NJ and UPGMA names are retained field labels; this example did not infer either tree by those methods.
Keep the source of each support value, its scale and its intended pairing. This route compares array positions. It does not automatically attach those positions to split identities. The displayed shared-split count is a separate topology count; matching array length to that count does not prove that the rows are correctly aligned.
2. Preserve paired positions and the declared scale
The valid case supplied [10, 20, 30] and [10, 30, 20]. The saved arrays and signed differences are shown below. Position is a zero-based array index, not a split label:
| Position | First value | Second value | Saved first − second |
|---|---|---|---|
0 | 10.0 | 10.0 | 0.0 |
1 | 20.0 | 30.0 | -10.0 |
2 | 30.0 | 20.0 | 10.0 |
The source displays support means with percent signs and compares the supplied numbers directly with 70 and 50. It does not multiply values by 100. For a percentage convention, supply and document that convention consistently; do not silently mix proportions on a 0–1 scale with percentages on a 0–100 scale. On the declared percent scale, the saved differences are percentage-point differences.
The reviewed code checks equal counts and finite values, but provides no automatic scale detection or 0–100 range-validation contract. The saved cases do not establish behavior for constant vectors, multidimensional inputs, infinity, out-of-range values or other invalid text. Do not treat a numeric conversion as validation of a support value’s scientific meaning.
3. Choose Split Support and read its complete summary
The displayed method label is Split Support; its saved table is called Split Support Comparison. All twenty routes initially start selected in the reviewed source. Each retained check selected this route alone, with zero calls to the other nineteen routes.
The saved raw correlation is 0.5; the earlier independent reference gives 1 / 2. No correlation was recomputed for this lesson. The complete successful summary is:
| Saved field | Saved value | Interpretation |
|---|---|---|
| Correlation | 0.5000 | Pearson correlation of the supplied paired arrays; formatted to four decimal places. |
| NJ Mean | 20.0% | Mean of the first supplied array, formatted to one decimal place with a percent sign. |
| UPGMA Mean | 20.0% | Mean of the second supplied array with the same formatting. |
| Shared Splits | 3 | Count from the parsed trees; it does not certify the support-row mapping. |
| Ref | Felsenstein, Evolution 1985 | Background citation, not evidence that bootstrap samples were generated in this run. |
A correlation of one half is not 50% confidence in either tree. The saved logs also report zero pairs where both values exceed 70 and three where both are below 50. Those are descriptive counts on this supplied scale, with strict inequalities. They do not calibrate a significance threshold.
Felsenstein’s bootstrap paper describes resampling data to assess phylogenetic uncertainty. This retained route receives already supplied support arrays; it does not generate the bootstrap samples or establish the provenance of those arrays.
4. Preserve the two complete refusal tables
With first-array values unchanged, the unequal-length case supplied [10, 30] as the second array. It was refused without truncation. A separate case supplied [10, "NaN", 20]; the quoted text converted to a nonfinite value at position 1 and was refused without dropping or replacing that row.
Each refusal produces its own three-field table. Read every field and keep it with the corresponding input:
| Saved case | Field | Saved value |
|---|---|---|
| Unequal lengths | Status | no_data |
| Unequal lengths | Required upload | nj_ |
| Unequal lengths | Ref | Felsenstein, Evolution 1985 |
| Quoted NaN | Status | no_data |
| Quoted NaN | Required upload | finite paired values in nj_ |
| Quoted NaN | Ref | Felsenstein, Evolution 1985 |
The route returned false before creating correlation or difference statistics in both cases. Each saved internal result event also says no_ and carries the same requirement. A refusal does not mean the correlation is zero. Preserve the input and reason; any revised pairing or missing-value policy needs an explicit decision before a later run.
5. Distinguish loaded data from a computed result
In both refused cases, the trees loaded successfully and the app’s has_ state remained true. The overall producer still ended its log with Complete. Those observations do not override the selected method’s no_ table.
Check the named method, its status and its requirement text. Do not infer successful support analysis from loaded trees, a general completion message or another method’s result. The retained calls used the input contract directly; they do not verify native loading, visual feedback or a hosted upload flow.
6. Keep numerical evidence and export scope separate
The saved observations preserve all input arrays, complete tables, full raw correlation and difference values, logs and refusal events. The successful table formats the correlation and means before storage; it is not a full-precision copy of every input or intermediate value.
These three checks performed zero native exports and no graphics. The website evidence download makes the retained observations reviewable, but it is not a native CSV or Export All package. A future export needs its own complete-file verification; another Tree method’s export receipt does not establish this route’s output completeness.