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:
rakeR has three main functions. The first stage is always to use
rk_weight() to produce a matrix of fractional weights. This matrix stores weights for each individual in each zone.
From this weight matrix,
rakeR has functions to create fractional weights (
rk_extract()) or integerised cases (
rk_integerise()), depending on your needs and use cases. Fractional (
rk_extract()ed) weights are generally more accurate, while integer cases are probably the most intuitive to use and are useful as inputs for further modeling.
To perform the weighting you should supply two data frames. One data frame should contain the constraint information (
cons) with counts per category for each zone (e.g. census counts). The other data frame should contain the individual–level data (
inds), i.e. one row per individual.
In addition, it is necessary to supply a character vector with the names of the constraint variables in
vars). This is so that
rakeR knows which are the contraint variables and which variables it should be simulating as an output.
Below are examples of
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 and the unique levels of the
age variable in
inds precisely match these:
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
rk_weight() which returns a data frame of raw weights.
weights <- rk_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
rk_integerise() them into a useable form.
rk_extract() produces aggregated totals of the simulated data for each category in each zone.
rk_extract()ed data is generally more accurate than
rk_integerise()d data, although the user should be careful this isn’t spurious precision based on context and knowledge of the domain. Because
rk_extract() creates a column for each level of each variable, numerical variables (e.g. income) must be removed or
cut() (otherwise the result would include a new column for each unique numerical value):
inds$income <- cut(inds$income, breaks = 2, include.lowest = TRUE, labels = c("low", "high")) ext_weights <- rk_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
rk_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.
rk_integerise() function produces a simulated data frame populated with simulated individuals. This is typically useful when:
int_weights <- rk_integerise(weights, inds = inds) int_weights[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
rk_integerise() returns one row per case, and the number of rows will match the known population (taken from
rk_rake() is a wrapper for
rk_weight() %>% rk_extract() or
rk_weight() %>% rk_integerise(). This is useful if the raw weights (from
rk_weight()) are not required. The desired output is specified with the
output argument, which can be specified with
"fraction" (the default) or
"integer". The function takes the following arguments in all cases:
Additional arguments are required depending on the output requested. For
output = "fraction":
output = "integer":
rake_int <- rk_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 <- rk_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.
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.
Copyright 2016-18 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/.