Estimate and Manage Empirical Distributions.

distrr

Overview

distrr provides some tools to estimate and manage empirical distributions. In particular, one of the main features of distrr is the creation of data cubes of estimated statistics, that include all the combinations of the variables of interest. The package makes strong usage of the tools provided by dplyr, which is a grammar of data manipulation.

The main functions to create a data cube are dcc5() and dcc6() (dcc stands for data cube creation).

The data cube creation is like:

data %>%
  group_by(some variables) %>%
  summarise(one or more estimated statistic)

in dplyr terms, but the operation is done for each possible combination of the variables used for grouping. The result will be a data frame in “tidy form”. See some examples in the Usage section below.

Installation

# From CRAN
install.packages("distrr")


# Or the development version from GitHub:
# install.packages("devtools")
devtools::install_github("gibonet/distrr")

Usage

Consider the invented_wages dataset:

library(distrr)
str(invented_wages)
#> Classes 'tbl_df' and 'data.frame':   1000 obs. of  5 variables:
#>  $ gender        : Factor w/ 2 levels "men","women": 1 2 1 2 1 1 1 2 2 2 ...
#>  $ sector        : Factor w/ 2 levels "secondary","tertiary": 2 1 2 2 1 1 2 1 2 1 ...
#>  $ education     : Factor w/ 3 levels "I","II","III": 3 2 2 2 2 1 3 1 2 2 ...
#>  $ wage          : num  8400 4200 5100 7400 4300 4900 5400 2900 4500 3000 ...
#>  $ sample_weights: num  105 32 36 12 21 46 79 113 34 32 ...

If we want to count the number of observations and estimate the average wage by gender, with dplyr we can do:

library(dplyr)
invented_wages %>%
  group_by(gender) %>%
  summarise(n = n(), av_wage = mean(wage))
#> # A tibble: 2 x 3
#>   gender     n av_wage
#>   <fct>  <int>   <dbl>
#> 1 men      547   5435.
#> 2 women    453   4441.

We can estimate the same statistics but grouped by education by changing the argument inside group_by:

invented_wages %>%
  group_by(education) %>%
  summarise(n = n(), av_wage = mean(wage))
#> # A tibble: 3 x 3
#>   education     n av_wage
#>   <fct>     <int>   <dbl>
#> 1 I           172   3774.
#> 2 II          719   5099.
#> 3 III         109   6139.

and estimate the statistics by gender and education including both variables in group_by:

invented_wages %>%
  group_by(gender, education) %>%
  summarise(n = n(), av_wage = mean(wage))
#> # A tibble: 6 x 4
#> # Groups:   gender [2]
#>   gender education     n av_wage
#>   <fct>  <fct>     <int>   <dbl>
#> 1 men    I            60   4627.
#> 2 men    II          409   5278.
#> 3 men    III          78   6886.
#> 4 women  I           112   3317.
#> 5 women  II          310   4865.
#> 6 women  III          31   4261.

With dcc5 we can perform all the steps above with one call:

invented_wages %>% 
  dcc5(.variables = c("gender", "education"), av_wage = ~mean(wage))
#> # A tibble: 12 x 4
#>    gender education     n av_wage
#>  * <fct>  <fct>     <int>   <dbl>
#>  1 Totale Totale     1000   4985.
#>  2 Totale I           172   3774.
#>  3 Totale II          719   5099.
#>  4 Totale III         109   6139.
#>  5 men    Totale      547   5435.
#>  6 men    I            60   4627.
#>  7 men    II          409   5278.
#>  8 men    III          78   6886.
#>  9 women  Totale      453   4441.
#> 10 women  I           112   3317.
#> 11 women  II          310   4865.
#> 12 women  III          31   4261.

The resulting data frame contains a column for each grouping variable, and the estimations of all the combinations of the variables:

• by gender
• by education
• by gender and education
• plus the same statistics for all the dataset, without any grouping (this can be set with the argument .all, which by default is TRUE).

Note that in the result there are some rows where the variables take the value "Totale". When a variable has this value, it means that the subset of the data considered in that row contains all the values of the variable. For example, the first row of the result of dcc5 contains the estimations for all the dataset. The value "Totale" can be changed with the argument .total.

The same result of dcc5 can be produced by dcc6, with a slightly different approach.

# Set a list of function calls
list_of_funs <- list(
  n = ~n(),
  av_wage = ~mean(wage),
  weighted_av_wage = ~weighted.mean(wage, sample_weights)
)

# Set the grouping variables
vars <- c("gender", "education")

# And create the data cube with dcc6
invented_wages %>% 
  dcc6(.variables = vars, .funs_list = list_of_funs, .total = "TOTAL")
#> # A tibble: 12 x 5
#>    gender education     n av_wage weighted_av_wage
#>  * <fct>  <fct>     <int>   <dbl>            <dbl>
#>  1 TOTAL  TOTAL      1000   4985.            4645.
#>  2 TOTAL  I           172   3774.            3527.
#>  3 TOTAL  II          719   5099.            4917.
#>  4 TOTAL  III         109   6139.            5885.
#>  5 men    TOTAL       547   5435.            5323.
#>  6 men    I            60   4627.            4681.
#>  7 men    II          409   5278.            5129.
#>  8 men    III          78   6886.            6173.
#>  9 women  TOTAL       453   4441.            3614.
#> 10 women  I           112   3317.            3227.
#> 11 women  II          310   4865.            4225.
#> 12 women  III          31   4261.            4388.

Compared to the results obtained with dcc5, we added the weighted average of wages and changed the "Totale" value to "TOTAL".