diyar

https://github.com/olisansonwu/diyar

Repository

README.Rmd

---
output: github_document
---

```{r setup, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
```

# diyar
[![CRAN version](http://www.r-pkg.org/badges/version/diyar)](https://cran.r-project.org/package=diyar)
[![CRAN RStudio mirror downloads](https://cranlogs.r-pkg.org/badges/diyar)](https://www.r-pkg.org/pkg/diyar)
[![Coverage status](https://codecov.io/gh/OlisaNsonwu/diyar/branch/master/graph/badge.svg)](https://app.codecov.io/github/OlisaNsonwu/diyar?branch=master)
[![Travis build status](https://api.travis-ci.com/OlisaNsonwu/diyar.svg?branch=master)](https://app.travis-ci.com/github/OlisaNsonwu/diyar)

## Installation
```{r eval = FALSE}
# Install the latest CRAN release 
install.packages("diyar")

# Or, install the development version from GitHub
install.packages("devtools")
devtools::install_github("OlisaNsonwu/diyar")
```

## Overview
`diyar` is an `R` package for linking records with shared characteristics.
The linked records represent an entity, which depending on the context of the analysis can be unique patients, 
infection episodes, overlapping periods of care, clusters or other occurrences as defined by a case definition. This makes it useful in ordinarily complex analyses such as record linkage,  
contact or network analyses e.t.c. 

The main functions are `links()`, `episodes()` and `partitions()`.
They are flexible in regards to how they compare records, as well as what are considered matches. 
Their functionality can sometimes overlap however, each is better suited to particular use cases: 

  + `links()` - link records with no relevance to an index record. For example, deterministic record linkage
  + `episodes()` - link records in relation to an index record. For example, contact and network analysis.  
  + `partitions()` - link records in relation to a fixed interval.

## links()
**Key features;**

  + multi-stage record linkage. Here, multiple linkage criteria are assessed in a specified order of priority.

```{r warning = FALSE}
library(diyar)
data(missing_staff_id)
dfr_stages <- missing_staff_id[c("age", "hair_colour", "branch_office")]
priority_order_1 <- c("hair_colour", "branch_office")
priority_order_2 <- c("branch_office", "hair_colour")

dfr_stages$id.1 <- links(criteria = as.list(dfr_stages[priority_order_1]))
dfr_stages$id.2 <- links(criteria = as.list(dfr_stages[priority_order_2]))
```
  + create and use complex rules for record matching. This is done with a `sub_criteria()`.
```{r warning = FALSE}
sub.cri.1 <- sub_criteria(
  hair.color = dfr_stages$hair_colour,
  age = dfr_stages$age,
  match_funcs = c(
    "exact" = exact_match,
    "age.range" = range_match)
)
last_word_wf <- function(x) tolower(gsub("^.* ", "", x))
last_word_cmp <- function(x, y) last_word_wf(x) == last_word_wf(y)
not_equal <- function(x, y) x != y
sub.cri.2 <- sub_criteria(
  dfr_stages$branch_office, 
  dfr_stages$age,
  match_funcs = c(
    "last.word" = last_word_cmp,
    "not.equal" = not_equal)
)
sub.cri.3 <- sub_criteria(sub.cri.1, sub.cri.2, operator = "and")
sub.cri.1
sub.cri.2
sub.cri.3
dfr_stages$id.3 <- links(
  criteria = "place_holder",
  sub_criteria = list("cr1" = sub.cri.3)
)
dfr_stages
```
There are variations of `links()` like `links_wf_probabilistic()` and `links_af_probabilistic()` for specific use cases such as probabilistic record linkage.

## episodes()
**Key features;**

  + link records within a specified period from an index record.
```{r warning = FALSE}
dfr_2 <- data.frame(date = as.Date("2020-01-01") + c(1:5, 10:15, 20:25))
dfr_2$id.1 <- episodes(
  date = dfr_2$date, case_length = 2,
  episodes_max = 1)
```
  + change the index record.
```{r warning = FALSE}
dfr_2$pref <- c(rep(2, 8), 1, rep(2, 8))
dfr_2$id.2 <- episodes(
  date = dfr_2$date, case_length = number_line(-2, 2),
  episodes_max = 1, 
  custom_sort = dfr_2$pref)
```
  + add a recurrence period 
```{r warning = FALSE}
dfr_2$id.3 <- episodes(
  date = dfr_2$date, case_length = number_line(-2, 2), 
  episode_type = "rolling", recurrence_length = 1,
  episodes_max = 1, rolls_max = 1)
```  
+ link overlapping periods
```{r warning = FALSE}
dfr_2$period <- number_line(dfr_2$date, dfr_2$date + 5)
dfr_2$id.4 <- episodes(
  date = dfr_2$period, case_length = index_window(dfr_2$period),
  episodes_max = 1)
dfr_2
```

There are variations of `episodes()` like `episodes_wf_splits()` for specific use cases such as more efficient handling of duplicate records.

## partitions()
**Key features;**

  + link all records within a specific periods in time 
```{r warining = FALSE}
dfr_3 <- dfr_2["date"]
dfr_3$id.1 <- partitions(
  date = dfr_3$date, 
  window = number_line(as.Date(c("2020-01-10", "2020-01-17")), 
                       as.Date(c("2020-01-12", "2020-01-24")))
  )
```
  + link all records within a splits of an interval 
```{r warining = FALSE}
dfr_3$id.2 <- partitions(date = dfr_3$date, by = 3, separate = TRUE) 
dfr_3$id.3 <- partitions(date = dfr_3$date, length.out = 3, separate = TRUE)
dfr_3
```

Find out more!
  
  + number_line and overlaps - `vignette("number_line")`
  + Introduction to epidemiological case definitions with diyar - `vignette("episodes")`
  + Introduction to record linkage with diyar - `vignette("links")`
  + Divvy up events with partitions - `vignette("panes")`
  
## Bugs and issues
Please report any bug or issues with using this package [here](https://github.com/OlisaNsonwu/diyar/issues).

Owner

GitHub Events

Committers

Last synced: over 2 years ago

Issues and Pull Requests

Last synced: 11 months ago

Packages