model_vcov: Compute model variance-covariance matrix
In ha-pu/supportR: Helpful supporters

Description Usage Arguments Details Value See Also Examples

View source: R/model_vcov.r

The function model_vcov computes a variance-covariance matrix for a linear regression model. The user can provide a cluster variable to cluster the variance-covariance matrix and can select to use bootstraping for clustering. The output can serve as input for the model_summary and plot_interaction functions.

1	model_vcov(model, type, var_cluster)

`model`	Linear regression model for which the variance-covariance matrix is to be computed, requires class "lm" or "glm"
`type`	Indicator which type of variance-covariance matrix is to be computed: 0 - no clustering, 1 - clustering using `var_cluster`, 2 - clustering using `var_cluster` with bootstraping
`var_cluster`	Vector, matrix, or data.frame containing the variables that are used for clustering

The option type == 0 uses vcov from the base package. When type == 1 is selected, the variance-covariance matrix is computed using the cluster.vcov command from the multiwayvcov package: multiwayvcov::cluster.vcov(model, var_cluster). For type == 2, the variance-covariance matrix is computed using the cluster.boot command from the multiwayvcov package: multiwayvcov::cluster.boot(model, var_cluster, R = 1000, boot_type = "wild", wild_type = function() sample(c(-1, 1), 1)).

A list of class ("mod_vcov" "list") containing two items:

`model`	The linear regression model supplied as input
`vcov_mat`	A variance-covariance matrix for the linear regression model supplied in the function

For output's attributes depend on the type selected in the function call:

`type == 0`	The attribute "type" is set to "standard"
`type == 1`	The attribute "type" is set to "robust", the attribute "var_cluster" takes the value supplied as variable `var_cluster`
`type == 2`	The attribute "type" is set to "robust_boot", the attribute "var_cluster" takes the value supplied as variable `var_cluster`

cluster.vcov, cluster.boot

  data <- supportR::create_data()

  mod1 <- lm(firm_value ~ profit + cogs + rnd + competition * board_size, data = data)
  model_vcov(model = mod1, type = 0, var_cluster = NULL)
  model_vcov(model = mod1, type = 1, var_cluster = data$country)

  mod2 <- glm(female_ceo ~ profit + cogs + rnd + ceo_age * board_size, data = data, family = "binomial")
  model_vcov(model = mod2, type = 2, var_cluster = data$industry)