8000 GitHub - giacomovagni/oaxacad · GitHub
[go: up one dir, main page]

Skip to content
giacomovagniPublic

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

1 watching

Forks

Repository files navigation

Tidy Output for Blinder-Oaxaca Decomposition

The oaxacad package is a simple package allowing to output results from a Blinder-Oaxaca decomposition. This package is a wrapper of the oaxaca package.

Install the package directly from github with

library(devtools)
# install_github("giacomovagni/oaxacad")

Dependencies

To use microecon package please load the following libraries.

library(oaxaca)
library(tidyverse)
library(broom)
library(gtools)
#
library(oaxacad)

Simulate Data for Decomposition

The function oaxaca_data allows a simple way to simulate data for testing the Oaxaca decomposition.

We have two groups (A and B, or 0 and 1), and an outcome $Y$.

Imagine that we want a simulated dataset with a gap that is only due to observed characteristics (X).

The function allows to set one variable for observed characteristics. The variable is normally distributed and the user can set the mean and the sd.

In the function oaxaca_data, we can set the mean value of characteristics for group A with xA_avr = 10 and the mean value for group B with xB_avr = 5.

The intercepts between the two groups are the same, interceptA_avr = 100, interceptB_avr = 100, and the coefficients (the regression’ $\beta$) are the same betaA = 5, betaB = 5.

output = oaxaca_data(interceptA_avr = 100, interceptB_avr = 100, betaA = 5, betaB = 5, xA_avr = 10, xB_avr = 5)

The function output a plot for the two groups

output$fig
## `geom_smooth()` using formula = 'y ~ x'

The $X$ axis shows the difference in observed characteristics (X), and on the $Y$ axis the gap in the outcome Y.

The simulated data can be found in output$dataframe


The function oaxaca_sim() takes the same inputs as the function oaxaca_data but provides all the computations of the decomposition internally, allowing easy simulations.

sim1 = oaxaca_sim(betaA = 5, betaB = 8, xA_avr = 10, xB_avr = 10)
sim2 = oaxaca_sim(betaA = 5, betaB = 5, xA_avr = 8, xB_avr = 10)

sim1$simulated_data$fig

sim2$simulated_data$fig

sim1$decomp_reg
## NULL
sim2$decomp_reg
## NULL

Using the Oaxaca function

We compute the decomposition using the following command

#
output = oaxaca_data(interceptA_avr = 100, interceptB_avr = 110, betaA = 10, betaB = 5, xA_avr = 10, xB_avr = 5)
# using the package oaxaca
output = oaxaca::oaxaca(Y ~ X | group, data=output$dataframe)

We can tidy several outputs from the oaxaca function

Decomp_simple(output)
##                     gap   Explained Unexplained % explained
## coef(explained) 66.5642 25.7405 *** 40.8237 ***    38.67019
Decomp_Regression(output)
##              variable      β1 β1(SE) β1(pvalue)      β2 β2(SE) β2(pvalue)
## 1         (Intercept) 100.965  0.714          0 109.716  0.450          0
## 2                   X   9.899  0.062          0   5.050  0.063          0
## 3       decomposition      NA     NA         NA      NA     NA         NA
## 4 decomposition (\\%)      NA     NA         NA      NA     NA         NA
##       X1    X2 X1-X2 ΔXβ [endowment] ΔβX [coef]   ΔβΔX      Σ
## 1  1.000 1.000 0.000           0.000     -8.751  0.000 -8.751
## 2 10.222 5.125 5.097          25.740     24.854 24.720 75.315
## 3     NA    NA    NA          25.740     16.103 24.720 66.564
## 4     NA    NA    NA           0.387      0.242  0.371  1.000
Decomp(output)
## Omega = 1

## [[1]]
##             coef(endowments) se(endowments) coef(coefficients) se(coefficients)
## (Intercept)             0.00            0.0              -8.75             0.76
## X                      25.74            1.3              24.85             0.94
##             coef(interaction) se(interaction)
## (Intercept)              0.00            0.00
## X                       24.72            1.23
## 
## [[2]]
##                  endowment pval_endow coefficients pval_coef interaction
## coef(endowments)     25.74        ***       16.103       ***       24.72
##                  pval_int
## coef(endowments)      ***
## 
## [[3]]
##             Coeff_GroupRef Coeff_Group1   diff pvalue
## (Intercept)        100.965      109.716 -8.751      0
## X                    9.899        5.050  4.850      0
## 
## [[4]]
##             Composition_GroupRef Composition_Group1  diff pvalue
## (Intercept)                1.000              1.000 0.000    NaN
## X                         10.222              5.125 5.097      0
## 
## [[5]]
##               Decomposition
## Group1             135.5958
## GroupRef           202.1599
## TotalChange         66.5642
## .                          
## Compositional       25.7405
## Unexplained         16.1033
## Interaction         24.7204
## 
## [[6]]
##   perc_explained
## 1          38.67

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages

0