outlier_view

Back to L1.5 | Browse all axes | Browse all options

Axis outlier_view on sub-layer L1_5_D_missing_outlier_audit (layer l1_5).

Sub-layer

L1_5_D_missing_outlier_audit

Axis metadata

  • Default: 'iqr_flag'

  • Sweepable: False

  • Status: operational

Operational status summary

  • Operational: 4 option(s)

  • Future: 0 option(s)

Options

iqr_flag – operational

IQR-rule outlier flag per series (Tukey 1977).

L1.5.D outlier visualisation iqr_flag.

This option configures the outlier_view axis on the L1_5_D_missing_outlier_audit sub-layer of L1.5; output is emitted under manifest.diagnostics/l1_5/L1_5_D_missing_outlier_audit/ alongside the other selected views.

When to use

Robust to non-Gaussian distributions; flags values outside [Q1 - 1.5·IQR, Q3 + 1.5·IQR].

References

  • macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’

  • Tukey (1977) ‘Exploratory Data Analysis’, Addison-Wesley.

Related options: zscore_flag, multi, none

Last reviewed 2026-05-05 by macroforecast author.

multi – operational

Produce both IQR and z-score outlier flags.

L1.5.D outlier visualisation multi.

This option configures the outlier_view axis on the L1_5_D_missing_outlier_audit sub-layer of L1.5; output is emitted under manifest.diagnostics/l1_5/L1_5_D_missing_outlier_audit/ alongside the other selected views.

When to use

Cross-checking outlier counts across criteria; agreement strengthens the flag.

References

  • macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’

Related options: iqr_flag, zscore_flag, none

Last reviewed 2026-05-05 by macroforecast author.

none – operational

Skip outlier flagging.

L1.5.D outlier visualisation none.

This option configures the outlier_view axis on the L1_5_D_missing_outlier_audit sub-layer of L1.5; output is emitted under manifest.diagnostics/l1_5/L1_5_D_missing_outlier_audit/ alongside the other selected views.

When to use

Pre-cleaned panels where L2.C will not run; reducing diagnostic surface.

References

  • macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’

Related options: iqr_flag, zscore_flag, multi

Last reviewed 2026-05-05 by macroforecast author.

zscore_flag – operational

|z-score| > 3 outlier flag per series.

L1.5.D outlier visualisation zscore_flag.

This option configures the outlier_view axis on the L1_5_D_missing_outlier_audit sub-layer of L1.5; output is emitted under manifest.diagnostics/l1_5/L1_5_D_missing_outlier_audit/ alongside the other selected views.

When to use

Cheaper than IQR; assumes approximate normality. The 3σ threshold maps to ~0.3% tail probability under normality.

References

  • macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’

Related options: iqr_flag, multi, none

Last reviewed 2026-05-05 by macroforecast author.