Package: fastglm 0.1.4

Standard GLMs | Sparse and big.matrix paths | Negative-binomial regression | Firth bias-reduced GLMs | Firth across all decomposition methods | Firth on sparse and streaming designs | Hurdle models | Zero-inflated models | Summary | Ordinal models | References

Logistic regression under separation | General GLM families | Binomial (logit, probit, cloglog) | Poisson (log, sqrt) | Gamma (log, inverse) | Gaussian (identity, log) | Inverse Gaussian (log) | Standard errors | Assessing convergence | Adjusted scores and penalized likelihood | Profile intervals and tests for logistic regression | Speed comparison across families | References

Sparse and mapped matrices | Preserving a model across chunks | A more stable solve for related predictors | Bias reduction with repeated block reads | Prediction and robust inference | Ordered outcomes from repeatable blocks | Robust covariance for ordered outcomes | Assessment and testing without collecting the observations | Parquet and database sources | Count and two-part models from the same readers

Ordinal functionality at a glance | Models and estimation | Inference, assessment, and computation | Models and estimation | Common slopes and the cumulative-logit model | Housing satisfaction: fitting and interpretation | Alternative cumulative links | Partial proportional odds and nonparallel effects | The range of valid predictions | Mean bias reduction with varying effects | Prediction and effect summaries | Continuous predictors and local changes | Hypothesis tests and uncertainty | Coefficient profiles and score tests | Comparing common and varying effects | Robust covariance and bootstrap uncertainty | Model assessment | Residuals and reporting | Prediction scores, calibration, and graphical slope comparisons | Testing the common-slope restriction | Separation and a bias-reduced null | Repeated fits and large-data computation | Simulating responses from a fitted model | Prepared designs and repeated outcomes | Streaming from repeatable sources | Assessment and common-slope tests on the original source | Numerical computation and likelihood conventions | Locating small probabilities and predictor gaps | Missing intermediate categories within a predictor pattern | Category support and limits of inference

Example | Computational stability

Data-generating process | Convergence comparison | Summary | Log-likelihood comparison | Theta estimates | Cases where only fastglm converges

Negative-binomial variation | Positive counts and hurdle models | Zero-inflated counts | Testing associations within and across components | Larger datasets and common model specifications | Uncertainty for average counts

Data-generating process | Scoring convergence | Running the benchmark | Results | References

Package scope and model selection | Models and estimation | Inference, prediction, and assessment | Data size and repeated fitting | Specifying the mean model | Choosing a decomposition | Inference and population averages | Joint coefficient tests | Repeated outcomes on a shared design

Fitting and interpreting the components | Marginal predictions | Comparing average outcomes | Reading large outcome datasets in blocks