Experiments / tool

Independent-day planning

Calculate confirmation workload with noncentral-t power and the pilot’s 80% upper SD bound.

Calculator and method

YOUR INPUTS

Numeric inputs appear in a shared link. CSV data are kept out of the link.

Model calculation · not experimental data
sᵤ = s√[(m−1)/χ²₀.₂₀,m−1] gap = −ln(1−Etrue)+ln(1−Emin)−2b Power = P(T > t₀.₉₇₅,N−1), with noncentrality gap√N/sᵤ
MODEL & BOUNDARIES

Know what the result means.

Noncentral-t calculation for an independent normal daily-mean model, using a one-sided 80% upper standard-deviation bound from fresh pilot days. The decision uses the lower end of a two-sided 95% interval and two residual-bias allowances. Select the first N from 24 to 48 with at least 80% modeled optical power. WX-FOG-001 v0.2 specifies six fresh pilot days; changing that input explores a revised planning scenario. Defaults are hypothetical. Pilot and confirmation are separate; the subtotal excludes qualification, earlier discovery, resets, losses and calendar scheduling. This does not calculate joint power for other product gates or freeze a study.

Version 2.0 · SI quantities are used internally, with displayed input units converted explicitly. Retain your assumptions with any exported result.

METHOD REFERENCES

Follow the definitions.

  1. weatherhill · WX-FOG-001 v0.2
  2. NIST · Uncertainty in standard deviation
  3. SciPy · Noncentral Student t distribution

References support the physical relationships or measurement context. They do not validate any proposed weatherhill hardware.

Budget the experiment the protocol describes.

The default 35 days means valid, independently prepared confirmation days. It uses a hypothetical pilot standard deviation of 0.15 from six days, a true optical reduction of 35%, a decision threshold of 20% and a qualified residual log-bias bound of 0.05. These are planning inputs, not measured Weatherhill results.

How the calculation works

  1. Use the pilot size and variance to form the protocol’s 80% one-sided upper bound on standard deviation. The default bound is 0.21915 log units.
  2. Subtract the optical decision threshold and the protocol’s two-bias planning allowance from the expected log effect.
  3. Calculate the probability of clearing the upper critical value of the two-sided 95% t interval, using a noncentral-t distribution.
  4. Select the first integer from 24 through 48 valid confirmation days reaching 80% modeled optical power. Report an infeasible plan if the gap is nonpositive or the cap cannot meet that target.

At the defaults, 34 days gives about 79.4% modeled power; 35 gives about 80.6%. Using raw pilot SD without its upper uncertainty bound gives 18 days; the old normal shortcut gave 16. Those shorter scenarios do not include the protocol’s conservative variance-planning rule.

What 35 days includes

Two matched pairs on each day make 70 pairs and 140 confirmation runs. Six fresh pilot days are additional: 41 pilot-plus-confirmation days is only a subtotal. Instrument qualification, earlier discovery, invalid runs, replacement days and calendar constraints are not included. A run uses a five-minute valid measurement window, with stabilization and washout outside it.

The calculation is exact under the stated model of independent, normally distributed daily log effects. Pilot uncertainty coverage and power are separate probabilities. Valid optical confirmation also requires the specified metrology, data and reliability conditions. Product advancement adds the energy and application gates. An 80% modeled optical power is not an 80% chance of producing a useful product.

Keep exploring.

Search weatherhill

Search papers, tools and the project.