6 Introduction to the tongfen package

The tongfen R package (von Bergmann 2021) facilitates making data on different geometries comparable.

TongFen (通分) means to convert two fractions to the least common denominator, typically in preparation for further manipulation like addition or subtraction. In English, that’s a mouthful and sounds complicated. But in Chinese there is a word for this, TongFen, which makes this process appear very simple.

When working with geospatial datasets we often want to compare data that is given on different regions. For example census data and election data. Or data from two different censuses. To properly compare this data we first need to convert it to a common geography. The process to do this is quite analogous to the process of TongFen for fractions, so we appropriate this term to give it a simple name. Using the tongfen package, preparing data on disparate geographies for comparison by converting them to a common geography is as easy as typing tongfen.

In particular, the package has a number of convenience functions to facilitate making Canadian census data comparable through time, making it easy to perform longitudinal analysis on fine geographies based on the Canadian Census. Essentially, the tongfen package creates a semi-custom tabulation based on Dissemination Block, Dissemination Area, or Census Tract geographies.

These semi-custom tabulations are created in three steps:

Create a correspondence table for geographies from different censuses. By default the official StatCan correspondence files are used for that, but these only exist back to 2001 when the current geographic system based on dissemination blocks and dissemination areas started. To go back further, when enumeration areas were the basic building block, we need to rely on geospatial matching of the areas to create a harmonized common geography.
Create metadata that contains information about how the census variables of interest can be aggregated in the case where geographies get joined. For example, if we are interested in the share of households in low income, we need to know what this share is based on in order to correctly aggregate it up. CensusMapper holds detailed metadata, so this process is automated.
Join geographies and aggregate census data as described in the correspondence table from Step 1 and the metadata in step 2.

The result of this process is a semi-custom tabulation of the data we want that is created on the fly, at the price of coming on a slightly coarser geography than the original input geographies in cases where geographies had to be joined to create the harmonized geography.

To install the package from CRAN use

Code

install.packages("tongfen")

--- title: "Introduction to the **tongfen** package" label: tongfen-intro --- ![](images/tongfen-sticker.png){#fig-tongfen-logo style="float:right;margin-left:3em;" width="250"} The [tongfen](https://mountainmath.github.io/tongfen/index.html) R package [@tongfen] facilitates making data on different geometries comparable. TongFen (通分) means to convert two fractions to the least common denominator, typically in preparation for further manipulation like addition or subtraction. In English, that's a mouthful and sounds complicated. But in Chinese there is a word for this, TongFen, which makes this process appear very simple. When working with geospatial datasets we often want to compare data that is given on different regions. For example census data and election data. Or data from two different censuses. To properly compare this data we first need to convert it to a common geography. The process to do this is quite analogous to the process of TongFen for fractions, so we appropriate this term to give it a simple name. Using the **tongfen** package, preparing data on disparate geographies for comparison by converting them to a common geography is as easy as typing `tongfen`. In particular, the package has a number of convenience functions to facilitate making Canadian census data comparable through time, making it easy to perform longitudinal analysis on fine geographies based on the Canadian Census. Essentially, the **tongfen** package creates a semi-custom tabulation based on Dissemination Block, Dissemination Area, or Census Tract geographies. These semi-custom tabulations are created in three steps: 1. Create a correspondence table for geographies from different censuses. By default the official StatCan correspondence files are used for that, but these only exist back to 2001 when the current geographic system based on *dissemination blocks* and *dissemination areas* started. To go back further, when *enumeration areas* were the basic building block, we need to rely on geospatial matching of the areas to create a harmonized common geography. 2. Create metadata that contains information about how the census variables of interest can be aggregated in the case where geographies get joined. For example, if we are interested in the share of households in low income, we need to know what this share is based on in order to correctly aggregate it up. CensusMapper holds detailed metadata, so this process is automated. 3. Join geographies and aggregate census data as described in the correspondence table from Step 1 and the metadata in step 2. The result of this process is a semi-custom tabulation of the data we want that is created on the fly, at the price of coming on a slightly coarser geography than the original input geographies in cases where geographies had to be joined to create the harmonized geography. To install the package from CRAN use ```{r eval = FALSE} install.packages("tongfen") ```