cluster_se: Map-reduce clustered standard errors with 'biglm'

In gregobad/biglm-mapper: biglm map reduce

Description

Map-reduce clustered standard errors with biglm

Usage

cluster_se(file_list, fitted_model, ...)

Arguments

file_list	Character vector of data file names. Must be in .rds format.
fitted_model	Fitted biglm object.
...	Additional arguments to pass to chunk_clusterse clustervar: character name of variable to cluster on levels_union: named list with factor variables to standardize

Details

Computation DOES NOT do small sample DOF adjustment (I assume if you're using biglm, the data is large). Also importantly, computation assumes no group overlap across files (all observations from each group are in one file).

Value

A list with elements

• xxinv: (X'X)^-1

• cluster_vcov: cluster-robust variance-covariance matrix estimate

• g_cov: inner matrix in cluster-robust vcov matrix

• e_hat: mean of squared residuals

Examples

cluster_se(data_files, m, clustervar="group")

gregobad/biglm-mapper documentation built on May 9, 2019, 1:09 a.m.