---

name: discrete-and-count-models description: "Use statsmodels discrete choice and count model APIs for Logit, Probit, MNLogit, OrderedModel, Poisson, NegativeBinomial, zero-inflated, hurdle, truncated, conditional models, marginal effects, and convergence troubleshooting." disable-model-invocation: true metadata: disco-role: operating license: BSD 3-Clause

Discrete and count models

Use this sub-skill for binary choice, multinomial choice, ordered outcomes, count models, zero-inflated/hurdle/truncated counts, conditional models, marginal effects, post-estimation prediction, and failures such as perfect separation or non-convergence.

Workflow

1. Identify the outcome type: binary, ordered categorical, unordered multinomial, nonnegative count, overdispersed count, zero-inflated/hurdle, truncated, or grouped conditional.
2. Build endog and exog, adding a constant for array APIs. Formula wrappers exist for common classes through statsmodels.formula.api.
3. Fit the simplest defensible model first; only then add overdispersion, inflation, truncation, or conditional structure.
4. Inspect convergence, parameter magnitudes, predicted probabilities/means, and marginal effects. For broad hypothesis tests or residual diagnostics, cross-route to statistical-tests-and-diagnostics.

Read or run

• Read references/api-reference.md for model classes, signatures, and result surfaces.
• Read references/workflows.md for binary, multinomial/ordered, and count recipes.
• Read references/troubleshooting.md for perfect prediction, separation, overdispersion, zero inflation, convergence, and prediction shape issues.
• Run scripts/smoke_discrete_models.py to check binary Logit and Poisson workflows on deterministic local data.

Boundaries

• Use linear-and-formula-models for GLM setup unless the task explicitly needs discrete-model-specific post-estimation, marginal effects, zero inflation, truncation, or choice-model classes.
• Use time-series-analysis for time-indexed discrete models only when the time-series structure is the core request.
• Use datasets-results-graphics for result export, tables, or plots after fitting.