correlation_method
Back to L1.5 | Browse all axes | Browse all options
Axis
correlation_methodon sub-layerL1_5_E_correlation_pre_cleaning(layerl1_5).
Sub-layer
L1_5_E_correlation_pre_cleaning
Axis metadata
Default:
'pearson'Sweepable: False
Status: operational
Operational status summary
Operational: 3 option(s)
Future: 0 option(s)
Options
kendall – operational
Kendall tau rank correlation.
L1.5.E correlation method kendall.
This option configures the correlation_method 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
Conservative rank measure; smaller variance than Spearman in small samples (n < 30).
References
macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’
Related options: pearson, spearman
Last reviewed 2026-05-05 by macroforecast author.
pearson – operational
Pearson product-moment correlation – linear association measure.
L1.5.E correlation method pearson.
This option configures the correlation_method 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
Default; assumes approximate normality of pairs and linear association.
References
macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’
Related options: spearman, kendall
Last reviewed 2026-05-05 by macroforecast author.
spearman – operational
Spearman rank correlation – monotonic-association measure.
L1.5.E correlation method spearman.
This option configures the correlation_method 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
Robust to outliers and non-normal marginals; preferred when pairs have heavy tails.
References
macroforecast design Part 4: ‘diagnostic layers default-off; non-blocking; produce JSON + matplotlib views attached to manifest.diagnostics/.’
Related options: pearson, kendall
Last reviewed 2026-05-05 by macroforecast author.