| Title: | Plot Positive and Negative Predictive Values for Medical Tests |
|---|---|
| Description: | Functions to plot and help understand positive and negative predictive values (PPV and NPV), and their relationship with sensitivity, specificity, and prevalence. See Akobeng, A.K. (2007) <doi:10.1111/j.1651-2227.2006.00180.x> for a theoretical overview of the technical concepts and Navarrete et al. (2015) for a practical explanation about the importance of their understanding <doi:10.3389/fpsyg.2015.01327>. |
| Authors: | Gorka Navarrete [aut, cre] |
| Maintainer: | Gorka Navarrete <[email protected]> |
| License: | CC0 |
| Version: | 0.4.2.9000 |
| Built: | 2024-11-08 04:44:16 UTC |
| Source: | https://github.com/gorkang/bayesianreasoning |
Given a FP and a desired PPV, what is the Minimum Prevalence of a Condition
min_possible_prevalence(Sensitivity = 95, FP_test = 1, min_PPV_desired = 90)min_possible_prevalence(Sensitivity = 95, FP_test = 1, min_PPV_desired = 90)
Sensitivity |
Sensitivity of the test: [0-100] |
FP_test |
False positive rate (1-Specificity): [0-100] |
min_PPV_desired |
Which PPV is what you consider the minimum to trust a positive result in the test: [0-100] |
A description showing the minimum necessary prevalence.
# Example 1 min_possible_prevalence(Sensitivity = 99.9, FP_test = .1, min_PPV_desired = 70) "To reach a PPV of 70 when using a test with 99.9 % Sensitivity and 0.1 % False Positive Rate, you need a prevalence of at least 1 out of 429" # Example 2 min_possible_prevalence(100, 0.1, 98) "To reach a PPV of 98 when using a test with 100 % Sensitivity and 0.1 % False Positive Rate, you need a prevalence of at least 1 out of 21"# Example 1 min_possible_prevalence(Sensitivity = 99.9, FP_test = .1, min_PPV_desired = 70) "To reach a PPV of 70 when using a test with 99.9 % Sensitivity and 0.1 % False Positive Rate, you need a prevalence of at least 1 out of 429" # Example 2 min_possible_prevalence(100, 0.1, 98) "To reach a PPV of 98 when using a test with 100 % Sensitivity and 0.1 % False Positive Rate, you need a prevalence of at least 1 out of 21"
plot_cutoff Create a cutoff plot, showing the healthy and sick distributions, and the consequences of different cutoff points
plot_cutoff( prevalence = 0.1, cutoff_point = 30, mean_sick = 35, mean_healthy = 20, sd_sick = 3, sd_healthy = 5, n_people = 1e+05, add_table = FALSE, output_filename = NULL )plot_cutoff( prevalence = 0.1, cutoff_point = 30, mean_sick = 35, mean_healthy = 20, sd_sick = 3, sd_healthy = 5, n_people = 1e+05, add_table = FALSE, output_filename = NULL )
prevalence |
prevalence of the disease |
cutoff_point |
cutoff point to use |
mean_sick |
mean for the sick people distribution |
mean_healthy |
mean for the healthy people distribution |
sd_sick |
sd for the sick people distribution |
sd_healthy |
sd for the healthy people distribution |
n_people |
number of people to use |
add_table |
FALSE/TRUE: add gt table with Sensitivity, Specificity, etc. |
output_filename |
NULL. If a filename, will save the plot |
A list with plots and table
## Not run: plot_cutoff(prevalence = 0.2) ## End(Not run)## Not run: plot_cutoff(prevalence = 0.2) ## End(Not run)
Plot PPV associated to different levels of FP and a specific Sensitivity, for two different Prevalence groups.
PPV_diagnostic_vs_screening( max_FP = 10, Sensitivity = 100, prevalence_screening_group = 100, prevalence_diagnostic_group = 2, labels_prevalence = c("Screening", "Diagnostic"), folder = "" )PPV_diagnostic_vs_screening( max_FP = 10, Sensitivity = 100, prevalence_screening_group = 100, prevalence_diagnostic_group = 2, labels_prevalence = c("Screening", "Diagnostic"), folder = "" )
max_FP |
False positive rate (1-Specificity) [0-100]. |
Sensitivity |
Sensitivity of the test [0-100]. |
prevalence_screening_group |
Prevalence of the screening group, 1 out of x [1-Inf]. |
prevalence_diagnostic_group |
Prevalence of the diagnostic group, 1 out of x [1-Inf]. |
labels_prevalence |
Labels to use for both groups. |
folder |
Where to save the plot (the filename would be automatically created using the plot parameters) |
Shows a plot or, if given a folder argument, saves a .png version of the plot
# Example 1 PPV_diagnostic_vs_screening( max_FP = 10, Sensitivity = 100, prevalence_screening_group = 1500, prevalence_diagnostic_group = 3 ) # Example 2. QWith custom labels PPV_diagnostic_vs_screening( max_FP = 10, Sensitivity = 100, prevalence_screening_group = 1667, prevalence_diagnostic_group = 44, labels_prevalence = c("20 y.o.", "50 y.o.") )# Example 1 PPV_diagnostic_vs_screening( max_FP = 10, Sensitivity = 100, prevalence_screening_group = 1500, prevalence_diagnostic_group = 3 ) # Example 2. QWith custom labels PPV_diagnostic_vs_screening( max_FP = 10, Sensitivity = 100, prevalence_screening_group = 1667, prevalence_diagnostic_group = 44, labels_prevalence = c("20 y.o.", "50 y.o.") )
Plot heatmaps showing the PPV for a given Sensitivity and a range of Prevalences and False Positive values or NPV values for a given Specificity and a range of Prevalences and True Positive values
PPV_heatmap( min_Prevalence = 1, max_Prevalence = 1000, Sensitivity = NULL, Specificity = NULL, limits_Sensitivity = NULL, limits_Specificity = NULL, one_out_of = FALSE, overlay = "no", overlay_labels = "", overlay_extra_info = FALSE, overlay_position_FP = NULL, overlay_position_FN = NULL, overlay_prevalence_1 = NULL, overlay_prevalence_2 = NULL, uncertainty_prevalence = "high", label_title = "", label_subtitle = "", Language = "en", folder = "", PPV_NPV = "PPV", steps_matrix = 100, DEBUG = FALSE, ... )PPV_heatmap( min_Prevalence = 1, max_Prevalence = 1000, Sensitivity = NULL, Specificity = NULL, limits_Sensitivity = NULL, limits_Specificity = NULL, one_out_of = FALSE, overlay = "no", overlay_labels = "", overlay_extra_info = FALSE, overlay_position_FP = NULL, overlay_position_FN = NULL, overlay_prevalence_1 = NULL, overlay_prevalence_2 = NULL, uncertainty_prevalence = "high", label_title = "", label_subtitle = "", Language = "en", folder = "", PPV_NPV = "PPV", steps_matrix = 100, DEBUG = FALSE, ... )
min_Prevalence |
[x] out of y prevalence of disease: [1-Inf] |
max_Prevalence |
x out of [y] prevalence of disease: [1-Inf] |
Sensitivity |
Sensitivity of test: [0-100] |
Specificity |
Specificity of test: [0-100] |
limits_Sensitivity |
c(min Sensitivity, max Sensitivity) |
limits_Specificity |
c(min Specificity, max Specificity) |
one_out_of |
Show y scale as 1 out of x [TRUE, FALSE] FALSE by default |
overlay |
Type of overlay: ["line", "area"] |
overlay_labels |
Labels for each point in the overlay. For example: c("80", "70", "60", "50", "40", "30", "20 y.o.") |
overlay_extra_info |
show extra info in overlay? [TRUE/FALSE] |
overlay_position_FP |
FP value (position in the x-axis) for each point in the overlay. For example: c(7, 8, 9, 12, 14, 14) |
overlay_position_FN |
FN value (position in the x-axis) for each point in the overlay. For example: c(7, 8, 9, 12, 14, 14) |
overlay_prevalence_1 |
Prevalence value (position in the y-axis) for each point in the overlay. For example: c(1, 1, 1, 2, 1, 1) |
overlay_prevalence_2 |
Prevalence value (position in the y-axis) for each point in the overlay. For example: c(26, 29, 44, 69, 227, 1667) |
uncertainty_prevalence |
How much certainty we have about the prevalence ["high"/"low"] |
label_title |
Title for the plot |
label_subtitle |
Subtitle for the plot |
Language |
Language for the plot labels: ["sp", "en"] |
folder |
Where to save the plot (the filename would be automatically created using the plot parameters) |
PPV_NPV |
Should show PPV or NPV ["PPV", "NPV"] |
steps_matrix |
width of PPV/NPV matrix. 100 by default |
DEBUG |
Shows debug warnings [TRUE/FALSE] |
... |
Other parameters. Now used to pass dpi, height and width in the Show and Save plot section |
Shows a plot or, if given a folder argument, saves a .png version of the plot
PPV_heatmap( min_Prevalence = 1, max_Prevalence = 1000, Sensitivity = 100, Specificity = 98, Language = "en" )PPV_heatmap( min_Prevalence = 1, max_Prevalence = 1000, Sensitivity = 100, Specificity = 98, Language = "en" )
remove_layers_cutoff_plot Remove layers from a cutoff plot. This is useful to show how different things are calculated (e.g. Sensitivity)
remove_layers_cutoff_plot(cutoff_plot, delete_what, silent = TRUE)remove_layers_cutoff_plot(cutoff_plot, delete_what, silent = TRUE)
cutoff_plot |
A plot_cutoff() plot |
delete_what |
Elements to delete (i.e. FP, FN, TP, TN) |
silent |
TRUE do not show debug info |
a cutoff plot without the elements deleted
## Not run: PLOT = plot_cutoff(prevalence = 0.2) remove_layers_cutoff_plot(PLOT$final_plot, delete_what = c("FN", "TP")) + ggplot2::labs(subtitle = "Specificity = TN/(TN+FP)") ## End(Not run)## Not run: PLOT = plot_cutoff(prevalence = 0.2) remove_layers_cutoff_plot(PLOT$final_plot, delete_what = c("FN", "TP")) + ggplot2::labs(subtitle = "Specificity = TN/(TN+FP)") ## End(Not run)