---

name: nullity-utilities description: "Guides missingno nullity_filter and nullity_sort DataFrame completeness filtering, sorting, examples, and troubleshooting." disable-model-invocation: true metadata: disco-role: operating license: MIT

missingno Nullity Utilities

Use this sub-skill when a task asks how to filter columns by completeness, sort rows or columns by missingness, or understand the filter, n, p, sort, and axis parameters used by missingno plots.

Core APIs

import missingno as msno

filtered = msno.nullity_filter(df, filter="bottom", n=10)
sorted_rows = msno.nullity_sort(df, sort="ascending", axis="columns")

Read references/api-reference.md for exact source-backed behavior and references/workflows.md for recipes.

Quick semantics

API	Use it for	Key behavior
nullity_filter(df, filter="top", p=.9, n=5)	Keep complete-enough columns	Threshold by non-null ratio, then cap to highest-completeness columns.
nullity_filter(df, filter="bottom", p=.5, n=10)	Keep incomplete columns	Threshold by non-null ratio, then cap to lowest-completeness columns.
nullity_sort(df, sort="ascending", axis="columns")	Sort rows by row completeness	axis="columns" counts across columns for each row.
nullity_sort(df, sort="descending", axis="rows")	Sort columns by column completeness	axis="rows" counts down rows for each column.

Plot integration

Plot functions reuse these utilities:

• matrix filters columns, then sorts rows with axis="columns".
• bar filters columns, then sorts columns with axis="rows".
• heatmap filters columns, then sorts columns with axis="rows" before dropping all-full/all-empty columns for correlation.
• dendrogram filters columns but does not sort before clustering.

For rendering and interpretation after filtering, read ../visualizations/SKILL.md.

Common decisions

1. Use filter="bottom" when the user wants the least complete columns or the variables most likely to reveal missingness problems.
2. Use filter="top" when the user wants mostly complete columns or wants to hide sparse variables.
3. Combine p and n when a percentage threshold and a maximum column count are both needed. Thresholding happens before the numeric cap.
4. Use axis="columns" to sort rows and axis="rows" to sort columns; the names follow pandas count-axis conventions, not natural-language target names.
5. Route plotting, labels, sparklines, heatmap interpretation, and dendrogram interpretation to ../visualizations/SKILL.md.

Troubleshooting

Read references/troubleshooting.md when:

• sort or axis raises ValueError.
• filter="top"/"bottom" produces a surprising column set.
• n and p are combined and the output contains fewer columns than expected.
• A plotting function receives an empty or too-small DataFrame after filtering.