MatchItEXT (version 0.0.1)

compute_res_var_ratio: Compute ratio of variances of residuals for covariates

Description

This function computes the ratio of variances of residuals for covariates, which was proposed by Rubin (2001). Applicable covariate types include continuous, binary and ordinal. Multinomial variables are not applicable to this function due to the absence of single residual. Usually a k-category multinomial variable will have k-1 residuals if multinomial logistic or probit regression is applied. For continuous variable, glm(family= gaussian) is used; for binary variable, glm(family= binomial(link= logit)) is used; for ordinal variable, MASS::polr(method = "logistic") is used, then single residual is obtained by using sure::resids().

Usage

compute_res_var_ratio(
  original_data = NULL,
  mi_obj = NULL,
  type_vec = NULL,
  discard = FALSE
)

Arguments

original_data

A data frame containing original data

mi_obj

A matchit object derived from MatchIt pacakge

type_vec

A vector specifying covariate types, valid values: 'ordinal', '3', 3; 'binary', '2', 2; 'continuous', '1', 1; 'excluded', '0', 0, NA. The last one means not to compute the ratio for this covariate. The length of this vector should be the same as that of covariate vector used in propensity score estimation.

discard

A logical value. TRUE means some observations are discarded before matching (with respect to discarded argument in matchit function), then the ratio before matching is based on the data after discard; FALSE means no observation is discarded before matching, then the ratio before matching is based on the original intact data.

References

Rubin, D. B. (2001). Using propensity scores to help design observational studies: Application to the tobacco litigation. Health Services and Outcomes Research Methodology, 2(3/4), 169-188. https://doi.org/10.1023/A:1020363010465

See Also

parse_formula() compute_var_ratio()

Examples

# NOT RUN {
 m_out <- MatchIt::matchit(treat ~ re74 + re75 + age + educ + hispan + black,
  data = MatchIt::lalonde, method = "nearest")
# use parse_formula() to check grouping variable and covariates
 parse_formula(m_out)
 compute_res_var_ratio(original_data = MatchIt::lalonde, mi_obj =
  m_out, type_vec = c(0, 1, 1, 1, 2, 2))
# }