Check the tetrad first, then the uncertainty calculation
The input covariances provide a direct way to calculate each tetrad. Changing the order of the indicators can change which covariance products are compared, so the reference must use the same order.
Bootstrap inference uses resampled data to assess uncertainty. It adds resampled tetrads, summary formulas and relationships among the restrictions tested together. This comparison checks the bias-corrected summary formulas without claiming that every underlying bootstrap model was independently fitted.
Which software produced these results?
These saved outputs come from the QuickPLS version listed below. The independent calculations use NumPy and SciPy. The version label identifies this test and does not describe a later rerun.
On a small screen, swipe the table sideways to see every column.
| Record | QuickPLS build or result checked | Reference | Comparison outcome |
|---|---|---|---|
| CTA tetrads and BC arithmetic | Saved 1.1.3 outputs | NumPy 1.26.4, SciPy 1.15.2 | Passed the comparisons documented for this record |
Tetrad values and bias-corrected interval arithmetic
The comparison includes 726 numerical or missing-result checks and 561 checks of output structure. The first group checks values and whether results are correctly marked as unavailable. The second checks the organisation of the output. These are different kinds of checks, not additional independent studies.
The calculations cover descriptive tetrads and bias-corrected summaries. Bias correction adjusts bootstrap interval limits to account for the position of the original estimate within the bootstrap distribution.
- Bias-corrected (BC) intervals differ from bias-corrected and accelerated (BCa) intervals. BCa adds an acceleration adjustment; a BC check does not establish that the BCa calculation has also been tested.
- The summary checks use saved resamples. They do not mean that every bootstrap model was fitted independently by the reference calculation.
| What was checked | Value / scope |
|---|---|
| Compared numbers and missing-value statuses | 726 |
| Outside the stated tolerance | 0 |
| Different scenario names in the comparison record | 1 |
| Largest absolute difference | 1.95536919989081e-10 |
Read the signed value and its absolute value
Each pair of rows shows a tetrad and its absolute value, which removes the sign. In the last pair, the tetrad is negative while its absolute value is positive. Both calculations preserve that distinction. A close numerical match says nothing by itself about whether the tetrad is statistically different from zero.
Values are displayed to 15 significant digits. Differences use the numbers before this display rounding, so equal-looking values can have a small nonzero difference. The download keeps the recorded precision and explains the reference definitions.
On a small screen, swipe the table sideways to see every column.
| Quantity compared | QuickPLS | Reference | Absolute difference | Comparison result |
|---|---|---|---|---|
Absolute tetrad (result row 1)Result identifierdescriptive_point/estimates/0/absolute_tetrad | 0.00697842664431114 | 0.00697842664431125 | 1.11022302462516e-16 | Pass |
Signed tetrad (result row 1)Result identifierdescriptive_point/estimates/0/tetrad | 0.00697842664431114 | 0.00697842664431125 | 1.11022302462516e-16 | Pass |
Absolute tetrad (result row 2)Result identifierdescriptive_point/estimates/1/absolute_tetrad | 0.00595356066186647 | 0.00595356066186625 | 2.22044604925031e-16 | Pass |
Signed tetrad (result row 2)Result identifierdescriptive_point/estimates/1/tetrad | 0.00595356066186647 | 0.00595356066186625 | 2.22044604925031e-16 | Pass |
Absolute tetrad (result row 3)Result identifierdescriptive_point/estimates/2/absolute_tetrad | 0.0129319873061776 | 0.0129319873061775 | 1.11022302462516e-16 | Pass |
Signed tetrad (result row 3)Result identifierdescriptive_point/estimates/2/tetrad | -0.0129319873061776 | -0.0129319873061775 | 1.11022302462516e-16 | Pass |
Error limits for tetrads and their summaries
Some outputs require exact agreement; others allow a small numerical difference. The table lists the rules used for each group of checks. Output-structure checks are separate from the numerical and missing-result checks in this table.
abs(native - aligned_reference) <= absolute_tolerance + relative_tolerance * abs(aligned_reference)
In this formula, native is the QuickPLS value and aligned_reference is the reference value expressed using the same definition. abs means absolute value, so the sign does not affect the size of the difference. Absolute tolerance is a fixed allowance. Relative tolerance adds an allowance that grows with the size of the reference value. A separate check verifies whether a result is available or missing where expected; it does not estimate another effect.
On a small screen, swipe the table sideways to see every column.
| Absolute tolerance | Relative tolerance | Numeric or missing-value checks | Where the comparison rule was specified |
|---|---|---|---|
| 1e-08 | 1e-08 | 10 | Specified separately for each compared value |
| 0 | 0 | 649 | Specified separately for each compared value |
| 1e-07 | 1e-07 | 67 | Specified separately for each compared value |
What these checks tell you
The evidence supports the checked tetrad calculations and bias-corrected summary arithmetic. It does not turn a nonsignificant test into confirmation that the measurement model is true.
Before interpreting a tetrad test, keep the indicator order and the set of hypotheses fixed. Also check the covariance definition and the rank rule, which determines how many restrictions are treated as independent. Several checked values can belong to the same analysis.
- Use the numerical limits shown above, and inspect missing-result handling and output structure separately. Agreement in a saved bootstrap summary is not the same as independently refitting all bootstrap samples.
Inspect the comparison files
The CSV contains selected tetrad comparisons. The JSON report records the versions, definitions, summary checks and output-structure checks. Use them together to distinguish a checked value from a checked uncertainty calculation.
To repeat the test, you also need the original input, indicator order, model, settings and software versions. The downloads report completed work and do not include the original datasets or application source code.
Download the comparison results
See the values we checked, the differences we found and the software versions used. The download also identifies the original test files.
About these tests
These are QuickPLS development comparisons. Each record identifies the reference software or mathematical calculation, the tested build and the settings used.
How to understand these comparisons · Find checks for another method