The error propagation module evaluates formulas on uncertain inputs and automatically propagates uncertainties.
- Data table: first row = headers, remaining rows = data
- Constants (optional): provided as text or a file; propagated together with the data
- Formula: an expression defined on the input variables
Two propagation methods are available:
- Taylor (derivative): fast approximation
- order=1: linear propagation (default)
- order=2: includes Hessian (second-derivative) contributions and applies a mean correction to the reported value (closer to Monte Carlo mean)
- Monte Carlo: samples each input from an independent normal distribution and returns “sample mean ± standard deviation”
- You can set the sample count (≥ 100) and an optional seed (reproducible runs)
The desktop app uses the same parsing rules as the computation core:
- Functions use Mathematica-style names and brackets (e.g.,
Sin[x],Log[x],Exp[x]) - Variables can be referenced by header names or supported aliases (see the in-app hints)
Use the function help button to view supported functions and examples.
After computation you will get:
- Per-row result value and combined uncertainty
- Uncertainty contribution plot (if enabled)
- LaTeX table (parentheses notation)
Note:
- Contribution breakdown is only available for Taylor modes (Monte Carlo does not return per-variable contributions).