Custom Distributions
There is no dedicated spreadsheet tab named custom_distributions in the current engine.
Instead, custom distributions are defined through one or more extra sheets referenced from the imp sheet.
How Custom Distributions Work
To define a custom distribution for a multiplier:
- In
imp, setDistribution = customfor that multiplier. - Create a separate sheet whose name matches the normalized
Impact Differentiator. - Put numeric values in the first column of that sheet.
- Optionally set
MinandMaxinimpto filter the empirical values before sampling.
Example:
In imp:
| Impact Differentiator | Min | Max | Distribution |
|---|---|---|---|
| PolicyAdditionality | 0.4 | 2.0 | custom |
Then create a sheet named PolicyAdditionality with values such as:
| A |
|---|
| 0.5 |
| 0.8 |
| 1.0 |
| 1.4 |
| 1.9 |
Important Naming Rule
The engine looks up custom-distribution sheets using the normalized multiplier name after removing spaces and dots.
Examples:
Policy Additionality-> sheetPolicyAdditionalityRisk.Reduction-> sheetRiskReduction
Parser Behavior
For each multiplier with Distribution = custom, the parser:
- reads the sheet with the normalized multiplier name,
- coerces the sheet to numeric values,
- uses only the first column,
- drops
NAvalues, - filters by
MinandMaxif those bounds are present, - stores the resulting empirical sample for later simulation.
During simulation, the engine samples from the empirical quantiles of that stored value vector.
Validation And Failure Modes
Current behavior:
- If the custom sheet cannot be read, parsing records an error.
- If the sheet exists but contains no usable numeric values after cleaning and filtering, downstream simulation is likely to fail or behave unexpectedly, so examples should always leave at least one valid numeric value.
Notes For Agents
- This page documents a pattern, not a literal sheet name.
- The current
Problem Definitionlanding page should not imply thatcustom_distributionsis a mandatory or optional tab name in the workbook. - If you generate example workbooks, keep the custom-sheet name aligned with the normalized differentiator name used internally by the parser.