old/old-D/README.md

In ttrodrigz/iterake: Tools for iterative raking

iterake

NOTE This package is under active development and is not yet ready to be used.

Overview

iterake performs an algorithm called iterative raking to generate row/respondent level weights for a sample of data. Iterative raking (also known as rim weighting), is one of several methods used to correct the deviation between the marginal proportions in your sample from a known population for a given set of variables. It is often useful to weight your data in order to make it as representative of the population as possible.

Three workhorse functions are inluded to help at every stage of the data weighting process: pre_rake() compares the unweighted data to the population model. iterake() performs the iterative raking. * post_rake() evaluates the effects of the weighting model.

Prior to weighting, two helper functions are used to construct the population model (your weighting targets): wgt_cat() constructs an individual weighting category (e.g., gender). pop_model() collects all of the individual weighting categories and prepares them for iterake().

This package was built to play nicely with the tidyverse, so the API should feel somewhat familiar to users of tidy tools.

Motivating Example

Suppose you've sampled 200 individuals from your home town to perform a study on used car buyers. In our omniscience, we know characteristics about age, gender, and the vehicle type owned in the population from which we are sampling.

AgeGenderVehicle Owned | Value | Pop. | Sample | |---------|:--------:|:--------:| | 18-34 | 32.0% | 34.0% | | 35-54 | 35.0% | 33.5% | | 55+ | 33.0% | 32.5% | | Value | Pop. | Sample | |---------|:--------:|:--------:| | Female | 54.0% | 57.0% | | Male | 46.0% | 43.0% | | Value | Pop. | Sample | |---------|:--------:|:--------:| | Car | 38.0% | 35.5% | | SUV | 47.0% | 51.5% | | Truck | 15.0% | 13.0% |

From the above tables, we can see there is a difference between what we collected in our sample and what is observed in the population. We can use the raking procedure to generate weights such that the weighted proportions of the sample will match (or come very close) to the population.

Note: we do not care about the joint probabilites of our weighting categories (i.e., percent of sample who are female, own an SUV, and are over 55 years old). As mentioned earlier, raking takes only into consideration the marginal probabilities.

Usage

This is where I realize we should be using RMarkdown/knitr so we can include code + output. Oops.

ttrodrigz/iterake documentation built on March 27, 2024, 12:48 a.m.