README.md

In marberts/ppd: Prices Production and Development

Prices Production and Development: An R package for working with price indices

On this page

• Description
• Installation
• Examples
• Changes
• References

Description

A collection of useful tools to simplify working with price indices in R.

Current features include:

• Functions to simplify working with monthly/quarterly/annual data.
• Functions to calculate geometric and harmonic averages.
• Functionality for calculating repeat-sales indices from the rsi package (Kirby-McGregor and Martin, 2019).
• Functions for calculating and aggregating price indices.
• Many examples for how to use this functionality for making and working with price indices.

These tools provide an easy way to produce simple, safe, and reusable code for working with price indices. The package is under active development; please get in touch if you have any ideas for features or code to contribute.

Installation

The most recent version of the ppd package can be installed by running

devtools::install_github("marberts/ppd")

Run install.packages("devtools") if the devtools packages is not installed.

The ppd package deliberately uses only functionality is base R and recommended packages (i.e., Matrix). No external libraries are needed, and this facilitates keeping track of versions for production code.

Examples

As an example of the functionality in ppd and the design philosophy behind the package, consider the problem of taking a monthly price index and turning it into a quarterly index by averaging the index values for each month in a quarter. Index data are stored in a dataset called data that looks like this:

|date |index| |--- |--- | |2017-01-01|100 | |2017-02-01|102 | |... |... | |2019-12-01|130 |

With ppd, the monthly index can easily be turned into a quarterly index as follows:

aggregate(index ~ year_quarter(date), FUN = mean, data = data, subset = complete_quarters(date))

Despite needing to know the names of the variables in data, this one-liner is very flexible. It automatically drops quarters that don't have an index value for all three months in that quarter, and works if a monthly series does not start or end on a quarter, or is missing index values for certain months. Dates can also be formatted inconsistently as the first day of the month, or the last day, or any day in between, and are returned by default as the first day of the month to be consistent with PRAD. All these features mean that the same snippet of code can be consistently and safely reused without having to know anything more than that the index is a monthly series.

As another example, consider calculating the quarter-over-quarter change in an index. As with the previous example, this can be easily done as:

data$qoq <- with(data, index / index[match(year_month(date, shift = -3), year_month(date))])

This approach to calculating a quarter-over-quarter change is quite robust; the index series does not need to be in chronological order, dates can be formatted inconsistently, there can be duplicate index values for a month, and the series can be missing index values for certain months. Very little needs to be known about the index series to safely calculate the quarter-over-quarter change.

As a final example, consider rebasing an index so that an entire year is the base period.

data$index <- with(data, 100 * index / mean(index[year(date) == '2017-01-01' & complete_years_m(date)]))

The subtlety with this example is the use of complete_years_m, and it ensures that a year must have an index value for every month to be a suitable base period. As with the other examples, this snippet of code can easily be reused because its use requires little prior knowledge of an index series.

Changelog

Changes in version 0.3.2

• Various improvements to documentation.
• Fixed a bug in pias_matrix where setting sparse = TRUE would always throw an error.

Changes in version 0.3.1

• Various improvements to documentation.
• Improved performance for generalized_mean with negative exponents.
• Fixed a bug in rs_z where NAs in inputs could produce incorrect output if the global na.action option is set to na.omit. An error is now thrown if any inputs have NAs.

Changes in version 0.3.0

• The packages has undergone major changes in this version that will likely break old code. The interface and implementation of most functions has changed to make them faster and safer to use.

References

Kirby-McGregor, M. and Martin, S. (2019). An R package for calculating repeat-sale price indices. Romanian Statistical Review, 3: 17-33.

marberts/ppd documentation built on March 27, 2020, 7:21 p.m.