ord_did: Ordinal Difference-in-Differences for Panel Data

In soichiroy/orddid: Difference-in-Differences Design for Ordinal Outcome

Description

ord_did() implements the difference-in-differences for the ordinal outcome.

Usage

ord_did(Ynew, Yold, treat, id_cluster = NULL, cut = c(0, 1),
  n_boot = 500, pre = FALSE, verbose = FALSE)

Arguments

Ynew	A numeric vector of ordinal outcome for the post-treatment period.
Yold	A numeric vector of ordinal outcome for the pre-treatment period.
treat	A numeric vector of treatment indicator. The treatment group should take 1 and the control group should take 0.
id_cluster	A vector of cluster id. If left as NULL, bootstrap is implemented at the individual level.
cut	A vector of cutoffs. Two numeric values should be specified. Default is cut = c(0, 1).
n_boot	The number of boostrapt iterations for estimating the variance. Default is n_boot = 500.
pre	A boolean argument used to indicate if the data comes entirely from pre-treatment periods. This should be TRUE when the output is supplied to equivalence_test.
verbose	If TRUE, print the progress of bootstrap iterations.

Value

ord_did() returns a list of class ‘orddid’ containing the following components:

fit	A list with the output of the ordinal DID estimators, which contains parameter estimates and predicted probabilities for each category.
boot	A list with the output of bootstraps, which contains parameter estimates and predicted probabilities for each category.
boot_params	A list with all objects generated during the bootstrap step.

Examples

## load packages
library(orddid)
library(dplyr)

## load example data
data("gun_twowave")

## run
## fit the ordinal DID
set.seed(1234)
fit <- ord_did(
  Ynew = gun_twowave %>% filter(year == 2012) %>% pull(guns),
  Yold = gun_twowave %>% filter(year == 2010) %>% pull(guns),
  treat = gun_twowave %>% filter(year == 2012) %>% pull(treat_100mi),
  id_cluster = gun_twowave %>% filter(year == 2010) %>% pull(reszip),
  n_boot = 10,
  pre = FALSE,
  verbose = FALSE
)

## view summary of the output
## non-cumulative effects
summary(fit, cumulative = FALSE)

## cumulative effects
summary(fit)

soichiroy/orddid documentation built on Oct. 3, 2020, 5:10 a.m.