QuickPLS Which kind of power is being checked?
Statistical power is the probability that a test detects a specified effect under the assumptions used in the calculation. This record concerns historical technical-power summaries, calculated from estimates already obtained for individual model paths.
The reference checks the normal-approximation formulas used for those summaries. That is useful arithmetic evidence, but it does not show how well a sample-size recommendation will work for a new study. These checks do not validate the current prospective procedure, which uses simulation before data collection.
Which saved output was compared?
The comparison uses saved QuickPLS output from the version listed below and an independent SciPy calculation. The original path model was not re-estimated for this check. Later product features and versions should therefore be assessed using their own evidence.
On a small screen, swipe the table sideways to see every column.
| Record | QuickPLS build or result checked | Reference | Comparison outcome |
|---|---|---|---|
| Technical power arithmetic | 1.01.15 | SciPy 1.15.2 | Passed the comparisons documented for this record |
Checking the saved path-level power calculations
The record contains 82 comparisons of output values and missing-result positions within the saved path-level normal-approximation results. These are individual checks within the example, not 82 separate paths or studies. The table shows the largest numerical difference and whether any check exceeded its allowed error.
- The comparison does not validate the separate study-wide posthoc procedure, which looks back at a completed study, or prospective power estimated by simulation for a future study.
| What was checked | Value / scope |
|---|---|
| Compared numbers and missing-value statuses | 82 |
| Outside the stated tolerance | 0 |
| Different scenario names in the comparison record | 1 |
| Largest absolute difference | 1.11022302462516e-16 |
Selected power values from the saved results
The labels identify a path-result row and its significance threshold. For the first path at α = 0.05, both displayed values are 0.481070168072413. This is retrospective power calculated from the saved model estimate, not a sample-size recommendation or a prospective assurance about a new study.
The table displays 15 significant digits. Differences use the values before this display rounding, so a very small difference may remain when the displayed pair looks equal. The download retains the original recorded precision and any adjustments used to compare the same calculation.
On a small screen, swipe the table sideways to see every column.
| Quantity compared | QuickPLS | Reference | Absolute difference | Comparison result |
|---|---|---|---|---|
Retrospective power at α = 0.01 (path row 1)Result identifierpath_technical_power_v1/0/alpha_01 | 0.233012417392554 | 0.233012417392554 | 2.77555756156289e-17 | Pass |
Retrospective power at α = 0.05 (path row 1)Result identifierpath_technical_power_v1/0/alpha_05 | 0.481070168072413 | 0.481070168072413 | 5.55111512312578e-17 | Pass |
Retrospective power at α = 0.01 (path row 2)Result identifierpath_technical_power_v1/1/alpha_01 | 0.779977695740689 | 0.779977695740689 | 0 | Pass |
Retrospective power at α = 0.05 (path row 2)Result identifierpath_technical_power_v1/1/alpha_05 | 0.926973064699504 | 0.926973064699505 | 1.11022302462516e-16 | Pass |
Retrospective power at α = 0.01 (path row 3)Result identifierpath_technical_power_v1/2/alpha_01 | 0.645451465340151 | 0.645451465340151 | 0 | Pass |
Retrospective power at α = 0.05 (path row 3)Result identifierpath_technical_power_v1/2/alpha_05 | 0.854187446323551 | 0.854187446323551 | 1.11022302462516e-16 | Pass |
The numerical rules used for the power comparison
abs(native - aligned_reference) <= absolute_tolerance + relative_tolerance * abs(aligned_reference)
In this formula, native is the saved QuickPLS value and aligned_reference is the independently calculated value under the same definition. The allowed difference combines a fixed amount and an amount proportional to the reference. The table separates decimal comparisons from exact checks. Expected missing results must agree as well.
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 | 30 | Specified separately for each compared value |
| 0 | 0 | 52 | Specified separately for each compared value |
What this tells you about sample-size planning
The small differences show agreement for the recorded arithmetic. They do not turn an observed-effect summary into a recommendation for recruiting participants. Planning requires a justified effect assumption and a design that matches the intended study.
For a simulation-based planning procedure, agreement also needs to be checked across simulated datasets, including failed fits and uncertainty in the estimated detection rate. Those questions are outside this historical comparison. The number of checked output values is not a count of independent studies.
- Use the error limits and full recorded values to judge this arithmetic comparison. Look for separate evidence when evaluating study-wide posthoc power or a prospective simulation procedure.
How to inspect the historical power files
The JSON report identifies the saved version, calculation settings and findings. The CSV provides the available comparison values. Match each row to its path-result position and significance setting before interpreting the number.
Repeating the comparison requires the original output or data, the same model and settings, and the corresponding software versions. These files report completed checks; the original datasets and app source code are not included.
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