gradient_value: Estimate the update gradient value

In distribglm: Distributed Generalized Linear Models

Description

Estimate the update gradient value

Usage

gradient_value(
  beta = NULL,
  data,
  formula,
  family = binomial(),
  iteration_number = 0,
  shuffle_rows = TRUE,
  link = NULL
)

use_glm_gradient_value(
  beta = NULL,
  data,
  formula,
  family = binomial(),
  iteration_number = 0,
  shuffle_rows = TRUE
)

Arguments

beta	current beta value, leave NULL to initialize
data	dataset to get gradient value from
formula	model formula to fit, with tilde syntax
family	generalized linear model family, see family
iteration_number	number of fitting iteration, used for tracking
shuffle_rows	should the rows of the dataset be permuted, so as to decrease privacy concerns
link	link function to use with family

Value

A list of estimated values, including the gradient, sample size, iteration number, covariance matrix (A_mat), number of samples with non-zero weights, the sum of the dispersion values (for overdispersion estimates), and a vector of values for combining to create the population gradient (u), with length of the number of beta values

Examples

data = data.frame(y = c(0, 0, 1),
pois_y = c(4, 1, 0),
x2 = c(-2.19021287072066,
       -0.344307138450805, 3.47215796952745),
x1 = c(-0.263859503846267,
       -0.985160029707486, 0.227262373184513))
gradient_value(data = data, formula = y ~ x1 + x2,
family = "binomial")
gradient_value(data = data, formula = pois_y ~ x1 + x2,
family = "poisson")
data = data.frame(y = c(0, 0, 1),
pois_y = c(4, 1, 0),
x2 = c(-2.19021287072066,
       -0.344307138450805, 3.47215796952745),
x1 = c(-0.263859503846267,
       -0.985160029707486, 0.227262373184513))
use_glm_gradient_value(data = data, formula = y ~ x1 + x2,
family = binomial(link = "probit"))

distribglm documentation built on April 15, 2021, 5:06 p.m.