README.md

In gerasy1987/usefulr: Various Useful Custom Functions

Something fascinating

Georgiy Syunyaev 2017-11-05

pacman::p_load(knitr)

opts_chunk$set(include = TRUE, 
               results = "markup", 
               message = FALSE, 
               warning = FALSE, 
               error   = FALSE)

usefulr::analyses() functionality

Load the package

pacman::p_load_gh("gerasy1987/usefulr")

Linear models (model = "lm")

• Here is the example of call to estimate and produce the output for lm (OLS) model with binary treatment, clustered SE’s and fixed effects. The output reports all coefficient estimates for treatment variable and no significance stars.

usefulr::analyses(DV = "gear",
                   treat = "vs",
                   model = "lm",
                   covs = NULL,
                   heterogenous = NULL,
                   cluster = "am",
                   subset = NULL,
                   FE = "carb",
                   status = c(T,T,F),
                   data = mtcars,
                   IPW = NULL,
                   treat_only = FALSE,
                   stars = FALSE)

## Call:
## usefulr::analyses(DV = "gear", treat = "vs", covs = NULL, heterogenous = NULL, 
##     subset = NULL, FE = "carb", cluster = "am", IPW = NULL, data = mtcars, 
##     model = "lm", treat_only = FALSE, status = c(T, T, F), stars = FALSE)
## 
## Estimation formula:
## [1] "gear ~ vs | carb | 0 | am"
## 
## Estimates:
##   term estimate std.error
## 1   vs    0.683     0.546
## 
## Summary:
## Adj. R2 = 0.280 , N = 32

• Here is the example of call to estimate and produce the output for lm (OLS) model with continous treatment on subsetted dataset with clustered SE’s and fixed effects. The output reports coefficient estimates for treatment variable and significance stars.

usefulr::analyses(DV = "mpg",
                  treat = "cyl",
                  model = "lm",
                  covs = "carb",
                  heterogenous = "hp",
                  cluster = "gear",
                  subset = "disp >= 0 & disp <= 300",
                  FE = "vs",
                  status = c(T,T,F),
                  data = mtcars,
                  treat_only = TRUE,
                  stars = TRUE)

## Call:
## usefulr::analyses(DV = "mpg", treat = "cyl", covs = "carb", heterogenous = "hp", 
##     subset = "disp >= 0 & disp <= 300", FE = "vs", cluster = "gear", 
##     data = mtcars, model = "lm", treat_only = TRUE, status = c(T, 
##         T, F), stars = TRUE)
## 
## Estimation formula:
## [1] "mpg ~ cyl + cyl:hp + hp + carb | vs | 0 | gear"
## 
## Estimates:
##     term estimate std.error
## 1    cyl   -5.172     0.933
## 2 cyl:hp    0.026     0.006
## 
## Summary:
## Adj. R2 = 0.621 , N = 21

Binary outcomes models (model = "logit" or model = "probit")

• Here is the example of call to estimate and produce the output for probit model with binary treatment, clustered SE’s. The output reports Adjusted McFadden R2, coefficient estimates only for treatment variable and no significance stars in printout.

usefulr::analyses(DV = "am",
                  treat = "vs",
                  model = "probit",
                  covs = "carb",
                  heterogenous = NULL,
                  cluster = "gear",
                  subset = NULL,
                  FE = NULL,
                  status = c(T,T,F),
                  data = mtcars,
                  IPW = NULL,
                  margin_at = NULL,
                  treat_only = TRUE,
                  stars = FALSE)

## Call:
## usefulr::analyses(DV = "am", treat = "vs", covs = "carb", heterogenous = NULL, 
##     subset = NULL, FE = NULL, cluster = "gear", IPW = NULL, data = mtcars, 
##     model = "probit", treat_only = TRUE, margin_at = NULL, status = c(T, 
##         T, F), stars = FALSE)
## 
## Estimation formula:
## [1] "am ~ vs + carb"
## 
## Estimates:
##   term estimate std.error
## 1   vs    0.788     0.963
## 
## Summary:
## Adj. R2 = -0.044 , N = 32

• Here is the example of call to estimate and produce the output for logit model with binary treatment on subsetted dataset with clustered SE’s and fixed effects. The output reports Adjusted McFadden R2, estimates and SE’s for average marginal effects for all independent variables (be careful with heterogenous marginal effects in non-linear models) and significance stars in printout.

usefulr::analyses(DV = "am",
                  treat = "vs",
                  model = "logit",
                  covs = NULL,
                  heterogenous = NULL,
                  cluster = "gear",
                  subset = "disp >= 0 & disp <= 300",
                  FE = "gear",
                  status = c(F,F,F),
                  data = mtcars,
                  margin_at = TRUE,
                  treat_only = FALSE,
                  IPW = NULL,
                  stars = TRUE)

## Call:
## usefulr::analyses(DV = "am", treat = "vs", covs = NULL, heterogenous = NULL, 
##     subset = "disp >= 0 & disp <= 300", FE = "gear", cluster = "gear", 
##     IPW = NULL, data = mtcars, model = "logit", treat_only = FALSE, 
##     margin_at = TRUE, status = c(F, F, F), stars = TRUE)
## 
## Estimation formula:
## [1] "am ~ vs + factor(gear)"
## 
## Estimates:
##   term estimate std.error
## 1   vs   -2.122     0.152
## 
## Summary:
## Adj. R2 = 0.330 , N = 21

Ordered categorical outcomes (model = "ologit" or model = "oprobit")

• Here is the example of call to estimate and produce the output for ologit (ordered logit) model with clustered SE’s and fixed effects. The output reports Adjusted McFadden R2, estimates and SE’s of coefficients for all independent variables and no significance stars.

usefulr::analyses(DV = "gear",
                  treat = "vs",
                  model = "ologit",
                  covs = NULL,
                  heterogenous = NULL,
                  cluster = "am",
                  subset = NULL,
                  FE = "am",
                  status = c(T,T,F),
                  data = mtcars,
                  margin_at = NULL,
                  IPW = NULL,
                  treat_only = FALSE,
                  stars = FALSE)

## Call:
## usefulr::analyses(DV = "gear", treat = "vs", covs = NULL, heterogenous = NULL, 
##     subset = NULL, FE = "am", cluster = "am", IPW = NULL, data = mtcars, 
##     model = "ologit", treat_only = FALSE, margin_at = NULL, status = c(T, 
##         T, F), stars = FALSE)
## 
## Estimation formula:
## [1] "gear ~ vs + factor(am)"
## 
## Estimates:
##   term estimate std.error
## 1   vs    0.552     2.794
## 
## Summary:
## Adj. R2 = 0.377 , N = 32

• Here is the example of call to estimate and produce the output for oprobit (ordered probit) model with clustered SE’s, fixed effects and McFadden R2. The output reports Adjusted McFadden R2, estimates of average marginal effects for treatment variable only on being in category "5" of the outcome and significance stars.

usefulr::analyses(DV = "gear",
                  treat = "vs",
                  model = "oprobit",
                  covs = NULL,
                  heterogenous = NULL,
                  cluster = "am",
                  subset = NULL,
                  FE = NULL,
                  status = c(T,T,F),
                  data = mtcars,
                  margin_at = "5",
                  IPW = NULL,
                  treat_only = TRUE,
                  stars = TRUE)

## Call:
## usefulr::analyses(DV = "gear", treat = "vs", covs = NULL, heterogenous = NULL, 
##     subset = NULL, FE = NULL, cluster = "am", IPW = NULL, data = mtcars, 
##     model = "oprobit", treat_only = TRUE, margin_at = "5", status = c(T, 
##         T, F), stars = TRUE)
## 
## Estimation formula:
## [1] "gear ~ vs"
## 
## Estimates:
##   term estimate std.error
## 1   vs    0.121     0.106
## 
## Summary:
## Adj. R2 = -0.005 , N = 32

gerasy1987/usefulr documentation built on Aug. 31, 2021, 4:25 p.m.