Skip to content

Normal loss model review checklist

A normal tail calculation needs both compatible parameter inputs and an explicit distribution assumption.

On this page

Retain the loss definition

Name the common unit, horizon, positive-loss sign and exposure or return transformation. Identify whether source dispersion is variance or standard deviation and whether a mean refers to return, profit or loss.

Check the distribution record

Keep a declared assumption separate from fitted parameters and normality evidence. Separate normal marginals do not establish a joint normal portfolio law, and a large sample alone does not establish normality of individual losses.

Review the measure and coefficient

Normal-model checks
LayerQuestion
ConfidenceWhich supported percentile?
VaRMean plus dispersion times quantile?
ESFirst moment divided by selected tail probability?
DensityHeight distinguished from interval area?
Zero scaleConstant limit used without standardization?
SupportFinite plot distinguished from unbounded law?

State any aggregation model

An independent additive-normal sum has separate mean and dispersion scaling. Other dependence, cash-flow or nonlinear setups need a different record. Retain unresolved evidence instead of treating a correct formula as validation of missing assumptions.

Further references