diff --git a/.Rbuildignore b/.Rbuildignore index eb41eaf..5a868c9 100644 --- a/.Rbuildignore +++ b/.Rbuildignore @@ -8,3 +8,4 @@ ^pkgdown$ ^CRAN-RELEASE$ ^CRAN-SUBMISSION$ +^README\.Rmd$ diff --git a/DESCRIPTION b/DESCRIPTION index 6601658..370cf7c 100644 --- a/DESCRIPTION +++ b/DESCRIPTION @@ -10,7 +10,11 @@ Authors@R: person(given = "Aurélien", family = "Allard", role = c("ctb"), - email = "aurelien.ab.allard@gmail.com")) + email = "aurelien.ab.allard@gmail.com"), + person(given = "Lukas", + family = "Jung", + role = c("ctb"), + email = "jung-lukas@gmx.net")) Description: The SPRITE algorithm creates possible distributions of discrete responses based on reported sample parameters, such as mean, standard deviation and range (Heathers et al., 2018, ). This package implements it, diff --git a/R/core-functions.R b/R/core-functions.R index 345f45c..49b933b 100644 --- a/R/core-functions.R +++ b/R/core-functions.R @@ -103,7 +103,7 @@ set_parameters <- function(mean, sd, n_obs, min_val, max_val, stop("The mean is outside the possible range, which is impossible - please check inputs.") } - if (!is.null(restrictions_minimum) && restrictions_minimum == "range") { + if (isTRUE(checkmate::check_choice(restrictions_minimum, "range"))) { restrictions_minimum <- list(1, 1) names(restrictions_minimum) <- c(min_val, max_val) } diff --git a/README.Rmd b/README.Rmd new file mode 100644 index 0000000..5ff7d3c --- /dev/null +++ b/README.Rmd @@ -0,0 +1,101 @@ +--- +output: github_document +--- + + + +```{r, include = FALSE} +knitr::opts_chunk$set( + collapse = TRUE, + comment = "#>", + fig.path = "man/figures/README-", + out.width = "100%" +) +``` + +# rsprite2 - Identify Distributions that Match Reported Sample Parameters + + + [![R-CMD-check](https://github.com/LukasWallrich/rsprite2/workflows/R-CMD-check/badge.svg)](https://github.com/LukasWallrich/rsprite2/actions) + [![Codecov test coverage](https://codecov.io/gh/LukasWallrich/rsprite2/branch/master/graph/badge.svg)](https://app.codecov.io/gh/LukasWallrich/rsprite2?branch=master) + [![R badge](https://img.shields.io/badge/Build%20with-♥%20and%20R-blue)](https://github.com/LukasWallrich/rsprite2) +[![R-CMD-check](https://github.com/LukasWallrich/rsprite2/workflows/R-CMD-check/badge.svg)](https://github.com/LukasWallrich/rsprite2/actions) + + +This package creates possible distributions based on reported sample parameters, using the SPRITE algorithm. This can be used to check what the original sample might have looked like, and thus to understand the data generation process better. This can help with the identification of fabricated data or misreported parameters, but also with checking whether sample characteristics such as floor/ceiling effects might suggest that findings are spurious. + +For that, it implements the SPRITE algorithm proposed by [Heathers et al. (2018)](https://peerj.com/preprints/26968/). Much of the code for is based on [Nick Brown's rSPRITE shiny app](https://github.com/sTeamTraen/rSPRITE), with some extensions. If you are just interested in interactive use and do not need these extensions (see advanced usage below) that Shiny online app (see https://shiny.ieis.tue.nl/sprite/) might be better for you. + +The package also includes dedicated functions to run the GRIM and GRIMMER tests. `GRIM_test()` checks whether a reported mean based on integers is consistent with a given sample size, while `GRIMMER_test()` checks whether a reported standard deviation is consistent with a reported mean and sample size. They can help catch impossible distributions and reporting errors without running any simulations. + +## Installation + +rsprite2 can be installed from CRAN with: + +```{r warning=FALSE, message=FALSE, eval=FALSE} +install.packages("rsprite2") +``` + +Or you might want to install the development version from GitHub: + +```{r, warning=FALSE, message=FALSE, eval=FALSE} +install.packages("remotes") +remotes::install_github('LukasWallrich/rsprite2') +``` + +## Simple example + +To create possible distributions, SPRITE needs the reported mean, standard deviation, sample size and range of values. + +For instance, what distribution of 20 responses on a 1 to 5 scale might result in a mean of 2.2 and a standard deviation of 1.3? + +```{r} +library(rsprite2) +sprite_parameters <- set_parameters(mean = 2.2, sd = 1.3, n_obs = 20, min_val = 1, max_val = 5) +find_possible_distribution(sprite_parameters) +``` + +If these means and standard deviations had been reported to two decimal places, they might be 2.20 and 1.33. NB: Given that trailing zeroes are ignored by R, you need to *specify the precision if mean or standard deviation ends in 0*. + +```{r} +sprite_parameters <- set_parameters(mean = 2.20, m_prec = 2, sd = 1.32, + n_obs = 20, min_val = 1, max_val = 5) +find_possible_distribution(sprite_parameters) +``` + +To find more than one distribution that matches the parameters, you can use `find_possible_distributions()` + +```{r} +find_possible_distributions(sprite_parameters, 10) +``` + +They can then be plotted to identify features that are shared across distributions. + +```{r} +res <- find_possible_distributions(sprite_parameters, 10) +plot_distributions(res) +``` + +## Advanced features + +### Multi-item scales + +Often, several Likert-type items are averaged to measure a single construct (e.g., participants might be asked how happy, satisfied and fulfilled they are, with the results averaged into a measure of well-being). Possible resulting distribution can be explored by rsprite2 by specifying the `n_items` parameter. + +```{r} +sprite_parameters <- set_parameters(mean = 1.95, sd = 1.55, n_obs = 20, + min_val = 1, max_val = 5, n_items = 3) +find_possible_distribution(sprite_parameters) +``` + +### Restrictions + +Sometimes, you might know how often certain values should appear in the distribution. Such restrictions need to be passed to the `set_parameters()` function as a named list. They can either specify the exact number of occurrences (`restrictions_exact`) or the minimum number of occurrences (`restrictions_minimum`) of a specific value. + +```{r} +sprite_parameters <- set_parameters(mean = 1.95, sd = 1.55, n_obs = 20, + min_val = 1, max_val = 5, n_items = 3, + restrictions_exact = list("3"=0, "3.67" = 2), + restrictions_minimum = list("1" = 1, "5" = 1)) +find_possible_distribution(sprite_parameters) +``` diff --git a/README.md b/README.md index 366baa5..b91d2f5 100644 --- a/README.md +++ b/README.md @@ -1,17 +1,40 @@ + + + # rsprite2 - Identify Distributions that Match Reported Sample Parameters - [![R-CMD-check](https://github.com/LukasWallrich/rsprite2/workflows/R-CMD-check/badge.svg)](https://github.com/LukasWallrich/rsprite2/actions) - [![Codecov test coverage](https://codecov.io/gh/LukasWallrich/rsprite2/branch/master/graph/badge.svg)](https://app.codecov.io/gh/LukasWallrich/rsprite2?branch=master) - [![R badge](https://img.shields.io/badge/Build%20with-♥%20and%20R-blue)](https://github.com/LukasWallrich/rsprite2) + +[![R-CMD-check](https://github.com/LukasWallrich/rsprite2/workflows/R-CMD-check/badge.svg)](https://github.com/LukasWallrich/rsprite2/actions) +[![Codecov test +coverage](https://codecov.io/gh/LukasWallrich/rsprite2/branch/master/graph/badge.svg)](https://app.codecov.io/gh/LukasWallrich/rsprite2?branch=master) +[![R +badge](https://img.shields.io/badge/Build%20with-♥%20and%20R-blue)](https://github.com/LukasWallrich/rsprite2) [![R-CMD-check](https://github.com/LukasWallrich/rsprite2/workflows/R-CMD-check/badge.svg)](https://github.com/LukasWallrich/rsprite2/actions) -This package creates possible distributions based on reported sample parameters, using the SPRITE algorithm. This can be used to check what the original sample might have looked like, and thus to understand the data generation process better. This can help with the identification of fabricated data or misreported parameters, but also with checking whether sample characteristics such as floor/ceiling effects might suggest that findings are spurious. - -For that, it implements the SPRITE algorithm proposed by [Heathers et al. (2018)](https://peerj.com/preprints/26968/). Much of the code for is based on [Nick Brown's rSPRITE shiny app](https://github.com/sTeamTraen/rSPRITE), with some extensions. If you are just interested in interactive use and do not need these extensions (see advanced usage below) that Shiny online app (see https://shiny.ieis.tue.nl/sprite/) might be better for you. - -The package also includes dedicated functions to run the GRIM and GRIMMER tests. `GRIM_test()` checks whether a reported mean based on integers is consistent with a given sample size, while `GRIMMER_test()` checks whether a reported standard deviation is consistent with a reported mean and sample size. They can help catch impossible distributions and reporting errors without running any simulations. +This package creates possible distributions based on reported sample +parameters, using the SPRITE algorithm. This can be used to check what +the original sample might have looked like, and thus to understand the +data generation process better. This can help with the identification of +fabricated data or misreported parameters, but also with checking +whether sample characteristics such as floor/ceiling effects might +suggest that findings are spurious. + +For that, it implements the SPRITE algorithm proposed by [Heathers et +al. (2018)](https://peerj.com/preprints/26968/). Much of the code for is +based on [Nick Brown’s rSPRITE shiny +app](https://github.com/sTeamTraen/rSPRITE), with some extensions. If +you are just interested in interactive use and do not need these +extensions (see advanced usage below) that Shiny online app (see +) might be better for you. + +The package also includes dedicated functions to run the GRIM and +GRIMMER tests. `GRIM_test()` checks whether a reported mean based on +integers is consistent with a given sample size, while `GRIMMER_test()` +checks whether a reported standard deviation is consistent with a +reported mean and sample size. They can help catch impossible +distributions and reporting errors without running any simulations. ## Installation @@ -30,52 +53,129 @@ remotes::install_github('LukasWallrich/rsprite2') ## Simple example -To create possible distributions, SPRITE needs the reported mean, standard deviation, sample size and range of values. +To create possible distributions, SPRITE needs the reported mean, +standard deviation, sample size and range of values. -For instance, what distribution of 20 responses on a 1 to 5 scale might result in a mean of 2.2 and a standard deviation of 1.3? +For instance, what distribution of 20 responses on a 1 to 5 scale might +result in a mean of 2.2 and a standard deviation of 1.3? ``` r library(rsprite2) sprite_parameters <- set_parameters(mean = 2.2, sd = 1.3, n_obs = 20, min_val = 1, max_val = 5) find_possible_distribution(sprite_parameters) +#> $outcome +#> [1] "success" +#> +#> $values +#> [1] 2 2 2 4 2 1 1 4 1 1 2 1 4 2 5 1 4 2 3 1 +#> +#> $mean +#> [1] 2.25 +#> +#> $sd +#> [1] 1.292692 +#> +#> $iterations +#> [1] 4 ``` -If these means and standard deviations had been reported to two decimal places, they might be 2.20 and 1.33. NB: Given that trailing zeroes are ignored by R, you need to *specify the precision if mean or standard deviation ends in 0*. +If these means and standard deviations had been reported to two decimal +places, they might be 2.20 and 1.33. NB: Given that trailing zeroes are +ignored by R, you need to *specify the precision if mean or standard +deviation ends in 0*. ``` r sprite_parameters <- set_parameters(mean = 2.20, m_prec = 2, sd = 1.32, n_obs = 20, min_val = 1, max_val = 5) find_possible_distribution(sprite_parameters) +#> $outcome +#> [1] "success" +#> +#> $values +#> [1] 4 4 1 1 1 1 1 4 1 2 2 4 4 2 1 2 1 1 4 3 +#> +#> $mean +#> [1] 2.2 +#> +#> $sd +#> [1] 1.321881 +#> +#> $iterations +#> [1] 4 ``` -To find more than one distribution that matches the parameters, you can use `find_possible_distributions()` +To find more than one distribution that matches the parameters, you can +use `find_possible_distributions()` ``` r find_possible_distributions(sprite_parameters, 10) +#> Only 9 matching distributions could be found. You can try again - given that SPRITE is based on random number generation, more distributions might be found then. +#> # A tibble: 9 × 6 +#> id outcome distribution mean sd iterations +#> +#> 1 1 success 2.2 1.32 8 +#> 2 2 success 2.2 1.32 2 +#> 3 3 success 2.2 1.32 3 +#> 4 4 success 2.2 1.32 4 +#> 5 5 success 2.2 1.32 4 +#> 6 6 success 2.2 1.32 3 +#> 7 7 success 2.2 1.32 1 +#> 8 8 success 2.2 1.32 1 +#> 9 9 success 2.2 1.32 4 ``` -They can then be plotted to identify features that are shared across distributions. +They can then be plotted to identify features that are shared across +distributions. ``` r res <- find_possible_distributions(sprite_parameters, 10) +#> Only 9 matching distributions could be found. You can try again - given that SPRITE is based on random number generation, more distributions might be found then. plot_distributions(res) -``` +#> Warning: Removed 18 rows containing missing values or values outside the scale range +#> (`geom_bar()`). +``` + + ## Advanced features ### Multi-item scales -Often, several Likert-type items are averaged to measure a single construct (e.g., participants might be asked how happy, satisfied and fulfilled they are, with the results averaged into a measure of well-being). Possible resulting distribution can be explored by rsprite2 by specifying the `n_items` parameter. +Often, several Likert-type items are averaged to measure a single +construct (e.g., participants might be asked how happy, satisfied and +fulfilled they are, with the results averaged into a measure of +well-being). Possible resulting distribution can be explored by rsprite2 +by specifying the `n_items` parameter. ``` r sprite_parameters <- set_parameters(mean = 1.95, sd = 1.55, n_obs = 20, min_val = 1, max_val = 5, n_items = 3) find_possible_distribution(sprite_parameters) +#> $outcome +#> [1] "success" +#> +#> $values +#> [1] 4.333333 1.000000 1.000000 1.000000 5.000000 1.000000 5.000000 4.333333 +#> [9] 1.000000 1.000000 1.000000 1.000000 3.000000 1.000000 3.333333 1.000000 +#> [17] 1.000000 1.000000 1.000000 1.000000 +#> +#> $mean +#> [1] 1.95 +#> +#> $sd +#> [1] 1.549476 +#> +#> $iterations +#> [1] 49 ``` ### Restrictions -Sometimes, you might know how often certain values should appear in the distribution. Such restrictions need to be passed to the `set_parameters()` function as a named list. They can either specify the exact number of occurrences (`restrictions_exact`) or the minimum number of occurrences (`restrictions_minimum`) of a specific value. +Sometimes, you might know how often certain values should appear in the +distribution. Such restrictions need to be passed to the +`set_parameters()` function as a named list. They can either specify the +exact number of occurrences (`restrictions_exact`) or the minimum number +of occurrences (`restrictions_minimum`) of a specific value. ``` r sprite_parameters <- set_parameters(mean = 1.95, sd = 1.55, n_obs = 20, @@ -83,4 +183,20 @@ sprite_parameters <- set_parameters(mean = 1.95, sd = 1.55, n_obs = 20, restrictions_exact = list("3"=0, "3.67" = 2), restrictions_minimum = list("1" = 1, "5" = 1)) find_possible_distribution(sprite_parameters) +#> $outcome +#> [1] "failure" +#> +#> $values +#> [1] 1.000000 1.000000 1.000000 1.000000 1.000000 1.000000 1.000000 4.000000 +#> [9] 1.000000 1.000000 1.000000 4.333333 1.000000 1.000000 1.000000 4.333333 +#> [17] 1.000000 5.000000 3.666667 3.666667 +#> +#> $mean +#> [1] 1.95 +#> +#> $sd +#> [1] 1.511264 +#> +#> $iterations +#> [1] 20000 ``` diff --git a/man/figures/README-unnamed-chunk-7-1.png b/man/figures/README-unnamed-chunk-7-1.png new file mode 100644 index 0000000..f9158ae Binary files /dev/null and b/man/figures/README-unnamed-chunk-7-1.png differ diff --git a/tests/testthat/test-core-and-plot-functions.R b/tests/testthat/test-core-and-plot-functions.R index 76fe20f..2e77b4c 100644 --- a/tests/testthat/test-core-and-plot-functions.R +++ b/tests/testthat/test-core-and-plot-functions.R @@ -26,6 +26,15 @@ test_that("Adjusting mean works", { expect_equal(sum(vec == 7), 0) }) +test_that("Example set_parameters() run works", { + expect_no_error( + set_parameters(mean = 1.95, sd = 1.55, n_obs = 20, + min_val = 1, max_val = 5, n_items = 3, + restrictions_exact = list("3"=0, "3.67" = 2), + restrictions_minimum = list("1" = 1, "5" = 1)) + ) +}) + restrictions <- list("2.33" = 1, "4" = 3, "4.67" = 0) parameters <- set_parameters(3.29, 1.29, 15, 1, 5, m_prec = 2, n_items = 3,