correlation_view
Back to L1.5 | Browse all axes | Browse all options
Axis
correlation_viewon sub-layerL1_5_E_correlation_pre_cleaning(layerl1_5).
Sub-layer
L1_5_E_correlation_pre_cleaning
Axis metadata
Default:
'none'Sweepable: False
Status: operational
Operational status summary
Operational: 4 option(s)
Future: 0 option(s)
Options
clustered_heatmap – operational
Clustered heatmap with hierarchical reorder of rows and columns.
L1.5.E correlation visualisation clustered_heatmap.
This option configures the correlation_view axis on the L1_5_E_correlation_pre_cleaning sub-layer of L1.5; output is emitted under manifest.diagnostics/l1_5/L1_5_E_correlation_pre_cleaning/ alongside the other selected views.
When to use
Large panels where cluster structure aids reading; reveals correlated variable blocks.
References
macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’
Related options: full_matrix, top_k_per_target, none
Last reviewed 2026-05-05 by macroforecast author.
full_matrix – operational
Full N×N correlation matrix as a heatmap.
L1.5.E correlation visualisation full_matrix.
This option configures the correlation_view axis on the L1_5_E_correlation_pre_cleaning sub-layer of L1.5; output is emitted under manifest.diagnostics/l1_5/L1_5_E_correlation_pre_cleaning/ alongside the other selected views.
When to use
Small panels (N < 50) where every pairwise correlation fits on one figure.
References
macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’
Related options: clustered_heatmap, top_k_per_target, none
Last reviewed 2026-05-05 by macroforecast author.
none – operational
Skip correlation diagnostics entirely.
L1.5.E correlation visualisation none.
This option configures the correlation_view axis on the L1_5_E_correlation_pre_cleaning sub-layer of L1.5; output is emitted under manifest.diagnostics/l1_5/L1_5_E_correlation_pre_cleaning/ alongside the other selected views.
When to use
Already covered by upstream EDA; reducing diagnostic surface.
References
macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’
Related options: full_matrix, clustered_heatmap, top_k_per_target
Last reviewed 2026-05-05 by macroforecast author.
top_k_per_target – operational
Top-k highest-|ρ| predictors per target.
L1.5.E correlation visualisation top_k_per_target.
This option configures the correlation_view axis on the L1_5_E_correlation_pre_cleaning sub-layer of L1.5; output is emitted under manifest.diagnostics/l1_5/L1_5_E_correlation_pre_cleaning/ alongside the other selected views.
When to use
Quickly identifying the most-correlated predictors when N is too large for a full matrix.
References
macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’
Related options: full_matrix, clustered_heatmap, none
Last reviewed 2026-05-05 by macroforecast author.