Create a spatial microsimulated data set in R using iterative proportional fitting (‘raking’).
Install the latest stable version from CRAN:
Or install the development version with
# Obtain devtools if you don't already have it installed # install.packages("devtools") # Install rakeR development version from GitHub devtools::install_github("philmikejones/rakeR")
Load the package with:
library("rakeR") #> #> Attaching package: 'rakeR' #> The following object is masked from 'package:stats': #> #> simulate
The overall process of microsimulating a data set is to
extract(). These return integer cases or fractional weights, respectively. Integer cases are probably the most intuitive to use and are useful as inputs for further modeling, but
extract()ed fractional weights can be more precise.
To perform the weighting you should supply two data frames, one with the constraint information (
cons) with counts per category for each zone (e.g. census counts) and one with individual–level data, i.e. one row per individual (
In addition, supply a character vector with the names of the constraint variables in
vars). This is so that any dependent variables included in
inds (for example
income below) are only used as output, not as part of the constraint process.
cons <- data.frame( "zone" = letters[1:3], "age_0_49" = c(8, 2, 7), "age_gt_50" = c(4, 8, 4), "sex_f" = c(6, 6, 8), "sex_m" = c(6, 4, 3), stringsAsFactors = FALSE ) inds <- data.frame( "id" = LETTERS[1:5], "age" = c("age_gt_50", "age_gt_50", "age_0_49", "age_gt_50", "age_0_49"), "sex" = c("sex_m", "sex_m", "sex_m", "sex_f", "sex_f"), "income" = c(2868, 2474, 2231, 3152, 2473) ) vars <- c("age", "sex")
It is essential that the unique levels in the constraint variables in the
inds data set match the variables names in the
cons data set. For example,
age_gt_50 are variable names in
cons. The unique levels of the
age variable in
inds precisely match these:
all.equal( levels(inds$age), colnames(cons[, 2:3]) ) #>  TRUE
Without this, the functions do not know how to match the
cons data and will fail so as not to return spurious results.
(Re-)weighting is done with
weight() which returns a data frame of raw weights.
weights <- weight(cons = cons, inds = inds, vars = vars) weights #> a b c #> A 1.227998 1.7250828 0.7250828 #> B 1.227998 1.7250828 0.7250828 #> C 3.544004 0.5498344 1.5498344 #> D 1.544004 4.5498344 2.5498344 #> E 4.455996 1.4501656 5.4501656
The raw weights tell you how frequently each individual (
E) should appear in each zone (
c). The raw weights are useful when validating and checking performance of the model, so it can be necessary to save these separately. They aren’t very useful for analysis however, so we can
extract() them into a useable form.
integerise() is useful when:
int_weights <- integerise(weights, inds = inds) int_weights[1:6, ] #> id age sex income zone #> 1 A age_gt_50 sex_m 2868 a #> 1.1 A age_gt_50 sex_m 2868 a #> 2 B age_gt_50 sex_m 2474 a #> 3 C age_0_49 sex_m 2231 a #> 3.1 C age_0_49 sex_m 2231 a #> 3.2 C age_0_49 sex_m 2231 a
integerise() returns one row per case, and the number of rows will match the known population (taken from
extract(), on the other hand, is more precise as the returned information is fractional, although the user should be careful this isn’t spurious precision based on context and knowledge of the domain.
extract() any numeric variables (such as income in the example) should be
cut() or removed.
inds$income <- cut(inds$income, breaks = 2, include.lowest = TRUE, labels = c("low", "high")) ext_weights <- extract(weights, inds = inds, id = "id") ext_weights #> code total age_0_49 age_gt_50 sex_f sex_m high low #> 1 a 12 8 4 6 6 2.772002 9.227998 #> 2 b 10 2 8 6 4 6.274917 3.725083 #> 3 c 11 7 4 8 3 3.274917 7.725083
extract() returns one row per zone, and the total of each category (for example female and male, or high and low income) will match the known population.
rake() is a wrapper for
weight() %>% integerise() or
weight() %>% extract(); useful if the raw weights are not required. The desired output is specified with the
output argument, which can be specified with
"fraction" (the default) or
integerise() respectively. The function takes the following arguments in all cases:
Additional arguments are required depending on the output requested. For
output = "fraction":
output = "integer":
Details of these context-specific arguments can be found in the respective documentation for
rake_int <- rake(cons, inds, vars, output = "integer", method = "trs", seed = 42) rake_int[1:6, ] #> id age sex income zone #> 1 A age_gt_50 sex_m high a #> 1.1 A age_gt_50 sex_m high a #> 2 B age_gt_50 sex_m low a #> 3 C age_0_49 sex_m low a #> 3.1 C age_0_49 sex_m low a #> 3.2 C age_0_49 sex_m low a
rake_frac <- rake(cons, inds, vars, output = "fraction", id = "id") rake_frac #> code total age_0_49 age_gt_50 sex_f sex_m high low #> 1 a 12 8 4 6 6 2.772002 9.227998 #> 2 b 10 2 8 6 4 6.274917 3.725083 #> 3 c 11 7 4 8 3 3.274917 7.725083
Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.
Feedback on the API, bug reports/issues, and pull requests are very welcome.
Do not push to
origin/master is a protected branch and expects CI tests to have been successfully completed before it will merge code.
Develop in a new branch, check your changes with
devtools::check(), and submit a pull request which will be checked by Travis CI.
Many of the functions in this package are based on code written by Robin Lovelace and Morgane Dumont for their book Spatial Microsimulation with R (2016), Chapman and Hall/CRC Press.
Their book is an excellent resource for learning about spatial microsimulation and understanding what’s going on under the hood of this package.
The book and code are licensed under the terms below:
Copyright (c) 2014 Robin Lovelace
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the “Software”), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
The rewighting (ipfp) algorithm is written by Andrew Blocker.
wrswoR package used for fast sampling without replacement is written by Kirill Müller.
Thanks to Tom Broomhead for his feedback on error handling and suggestions on function naming.
philmikejones at gmail dot com
Copyright 2016-17 Phil Mike Jones.
rakeR is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
rakeR is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with rakeR. If not, see http://www.gnu.org/licenses/.