Pulse Documentation

name: op-reg-glm
description: Generalized linear model via IRLS; supports binomial (logistic), poisson, and gamma families with family-specific link functions. Always buffered.
kind: operator
category: REG
operator: REG_GLM
type: reference
applies_to: process, compose, predict
examples_tags: [regression, glm, logistic, buffered-pipeline]

Regression operators emit coefficient + diagnostics; no Response.Components. Fit summaries ride Response.Regressions[i].

Use when

Fits a generalized linear model: logistic regression for yes/no outcomes, Poisson for counts, gamma for positive skewed amounts.

Questions it answers:

• Which customer traits are associated with the chance of churning, holding the others fixed?
• How does the expected number of support tickets change with account size?

Use something else:

• TEST_PROP_Z when you only want to know whether a yes/no rate differs between two groups.

Params

resample / selection are top-level modifiers — see op-reg-mod-resample, op-reg-mod-selection.

Inputs

target / predictors — numeric analytics set: u4/u8/u16/u32/u64, f32/f64, decimal128, date, packed_bool, plus nullable variants. Binomial targets accept {0,1} or packed_bool. Nullable rows drop from n_obs.

Output

RegressionResult: Coefficients["(intercept)"] + per-predictor βs (link scale); StdErrors, PValues (Wald-z from Cov(β) = (XᵀWX)⁻¹); CILower / CIUpper = Wald β ∓ z·SE (R confint.default, not the profile confint), ConfLevel, no DF; Deviance, NullDeviance, PseudoR2 = 1 − D/D₀ (McFadden only for 0/1 binomial); Family, Link echoed; ConvergedIters = IRLS steps; NObs raw rows, weighted adds SumWeights (+ NEff, probability); a robust covariance adds the Covariance echo. Always buffered — IRLS needs multiple passes.

Reading the output

• coefficients.*: The change in the linear predictor, on the link scale, for a one-unit increase in this predictor, holding the other predictors fixed: log-odds for binomial (logit link), the log of the expected count for Poisson (log link), the inverse of the expected value for gamma (inverse link).
• std_errors.*: How much the coefficient would vary from sample to sample, from the inverse of the fitted information matrix, with the dispersion fixed at 1.
• conf_level: The coverage ci_lower and ci_upper were built at (the request’s conf_level, 0.95 by default).

Gotchas

• Weighted: resample / selection → PULSE_WEIGHT_UNSUPPORTED; n_eff < p+1 → PULSE_WEIGHT_LOW_NEFF.
• penalty / alpha / l1_ratio rejected → PROCESSING_CONFIG. Regularized GLM is a later phase.
• Binomial separation diverges → PROCESSING_REGRESSION_NO_CONVERGE. Raise max_iters, drop the offender, or pre-bin.
• Unsupported link → PROCESSING_REGRESSION_INVALID_LINK.
• Dispersion fixed at 1 for binomial / poisson, and gamma (weighted too) inherits it — be sceptical of gamma SEs and overdispersed counts; covariance hc0–hc3 is dispersion-free.
• Robust covariance beside resample / selection, < 2 clusters or a repeated order_by → PROCESSING_REGRESSION_VCOV_UNSUPPORTED. A null cluster / order_by drops the row.
• Collinearity → PROCESSING_REGRESSION_RANK_DEFICIENT; drop a predictor.

See

• pulse_examples_search tags=[glm]
• Skills: regression-modeling, op-reg-ols, op-reg-bayes-linear, op-reg-mod-resample, op-reg-mod-selection