Meghansaha / affirm

Data affirmation and validation

affirm

The {affirm} package makes daily affirmation against our data. We use this package to affirm our raw data is as expected, and our derived variables continue to be accurate as data is updated.

1. Raw EDC Data: Use the affirmation functions to make assertions about the raw EDC data. Report errors back to data management, and in some cases make ‘corrections’ to the data. The issue will still be reported to DM, but these small corrections let us continue our work without dealing with data that should not be present in the data base and simplifies much of our logic downstream.

2. Derived Variables: Each time we derive a new variable, use data affirmation functions to affirm the derivation continues to be accurate as the data updates. We’ll often use the affirm_true(error=TRUE) in this setting, which throws an error when our underlying assumptions are not true and will require us to address the issue before continuing.

The {pointblank} is another package that performs data validations. {pointblank} is far more comprehensive than {affirm}, and {affirm} utilizes many of the ideas and reporting introduced in {pointblank} with defaults and reports tailored for the needs of the PCCTC organization. If {affirm} is not a perfect match for you, {pointblank} will likely meet all your validation needs!

Installation

1. Install the most recent release of {affirm} with

   devtools::install_github("pcctc/affirm@*release")

2. Install from a development branch for testing

   devtools::install_github("pcctc/affirm")

Examples

Load the package and initialize a new affirmation session with affirm_init()

library(affirm)
#> 
#> Attaching package: 'affirm'
#> The following object is masked from 'package:stats':
#> 
#>     filter

# initiate an affirmation session
affirm_init(replace = TRUE)
#> ✔ We're ready to make data affirmations...

Run the individual affirmations…

as_tibble(mtcars) |>
  affirm_true(
    label = "No. cylinders must be 4, 6, or 8",
    condition = cyl %in% c(4, 6, 8),
    id = 1,
    data_frames = "mtcars"
  ) |>
  affirm_true(
    label = "MPG should be less than 33",
    condition = mpg < 33,
    id = 2,
    data_frames = "mtcars"
  )
#> • No. cylinders must be 4, 6, or 8
#>   0 issues identified.
#> • MPG should be less than 33
#>   1 issue identified.
#> # A tibble: 32 × 11
#>     mpg   cyl  disp    hp  drat    wt  qsec    vs    am  gear  carb
#>   <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1  21       6   160   110  3.9   2.62  16.5     0     1     4     4
#> 2  21       6   160   110  3.9   2.88  17.0     0     1     4     4
#> 3  22.8     4   108    93  3.85  2.32  18.6     1     1     4     1
#> # ℹ 29 more rows

Create a report of the results

affirm_report_gt()

About the hex logo

The hex art is a depiction of Plato’s allegory of the cave. Each data table has an ideal form, and the {affirm} package will help us confirm that we are viewing it. Also, “forms”…get it? 😜