edhec-corr-heatmap

name	edhec-corr-heatmap
description	Produce a publication-ready correlation heatmap (PDF+PNG) from a CSV or in-memory dataframe. Triggers when the user asks for "correlation heatmap", "corr matrix figure", or uploads a *.csv with numeric columns and asks for a heatmap.
args	{"input":"path to CSV or name of in-memory dataframe","output":"optional stem for the figure (default figures/corr_heatmap) -- writes <stem>.pdf and <stem>.png","method":"pearson | spearman | kendall (default pearson)"}

Correlation heatmap

A four-phase skill. Each phase ends with a checkpoint the user can interrupt.

Phase 1 -- discover

Read the input. Identify numeric columns. For each column report:

• N non-missing
• % missing
• dtype

Stop and surface anything with > 5% missing, constant values (zero variance), or non-numeric dtype that survived coercion. Note that pairwise-complete observations are used by default (the pandas.corr default).

Phase 2 -- propose

Print:

• Final column list (after dropping all-NaN, constant, and non-numeric columns)
• Correlation method (default pearson)
• Output stem (default figures/corr_heatmap)

If the proposed matrix would have > 15 columns, stop and ask for confirmation.

Phase 3 -- compute

Compute the correlation matrix. Use:

• df.corr(method=method, numeric_only=True)
• Strict upper-triangle mask via np.triu(np.ones_like(corr, dtype=bool), k=1) so the diagonal is hidden too
• Round annotations to 2 decimals

Phase 4 -- write

Emit a heatmap figure:

• seaborn.heatmap(corr, mask=mask, annot=True, fmt=".2f", cmap="RdBu_r", vmin=-1, vmax=1, square=True, cbar_kws={"shrink": 0.7})
• plt.tight_layout()
• Save both <stem>.pdf (vector, for LaTeX) and <stem>.png (dpi=200, for previews)
• Default stem: figures/corr_heatmap
• Print both paths; do not commit unless asked