Loading...
Publication
Um novo pacote no R para visualizar dados em análise de sobrevivência
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 1.28 MB | Adobe PDF |
Authors
Abstract(s)
Existe uma plenitude de funcionalidades gr´aficas presentes no ecossistema
de pacotes da linguagem de programa¸c˜ao R para a ´area da Estat´ıstica e no
ramo da An´alise de Sobrevivˆencia, de tal modo que um utilizador pode n˜ao
saber onde come¸car tendo presente uma estrutura de dados ou um modelo
espec´ıfico. Neste trabalho detalha-se a funcionalidade de um novo pacote,
vsd, desenvolvido no R para visualizar dados em An´alise de Sobrevivˆencia
para suavizar esta problem´atica.
Come¸ca-se com uma introdu¸c˜ao ao processo de cria¸c˜ao de pacotes para
o R, com foco na cria¸c˜ao do esqueleto de um pacote atrav´es do pacote
devtools, na documenta¸c˜ao de um pacote e consequente declara¸c˜ao de funci onalidades atrav´es do pacote roxygen2, nalgumas componentes extras como
o uso de linting para corre¸c˜ao do c´odigo e a publica¸c˜ao do pacote, e o uso do
sistema de controlo de vers˜oes do c´odigo Git para f´acil publica¸c˜ao do pacote
sob licen¸ca aberta, como requerido pelo CRAN.
Em seguida, faz-se uma breve abordagem a conceitos em An´alise de So brevivˆencia relevantes para o pacote vsd.
O pacote vsd, que aproveita funcionalidades doutros pacotes j´a presen tes no ecossistema do R como os pacotes survminer e flexsurv, permite
visualizar v´arios gr´aficos para dados contendo censura `a direita como, por
exemplo: da estimativa de Kaplan-Meier da fun¸c˜ao de sobrevivˆencia, da esti mativa suavizada da fun¸c˜ao de risco, dos res´ıduos padronizados e dos gr´aficos
de floresta associados aos coeficientes do modelo de Cox, bem como das es timativas de modelos param´etricos. O pacote vsd foi constru´ıdo com um
design modular para permitir funcionalidade mais extensiva no futuro.
Os gr´aficos criados pelo pacote vsd podem ser explorados interativamente,
atrav´es do modo interativo com recurso opcional ao pacote plotly, ou ap´os
gera¸c˜ao modificados pelo utilizador, devido ao uso do pacote gr´afico ggplot2,
que permite reutiliza¸c˜ao e customiza¸c˜ao f´acil das estruturas gr´aficas criadas.
There’s a plethora of graphical packages on the R language ecosystem, specif ically in the area of Statistics and within the field of Survival Analysis, so much so that an user might not know where to start when considering a specific model or data structure. This dissertation details the functionality of a newly developed R package, vsd, meant to visualize data in Survival Analysis to ease this problem. It starts with an introduction on how to create R packages, with a focus on: building a package’s skeleton through the devtools package, in its docu mentation and consequent added functionality though the roxygen2 package, in some extra developmental concepts like linting for code correctness and package publication, and finally in how to use the version control system Git for easy availability of the package under an open license, as required by CRAN. Afterwards the discussion moves to a brief overview of any relevant con cepts under Survival Analysis for the developed package vsd. The vsd package, taking use of other packages on R’s ecosystem such as the packages survminer and flexsurv, allows the user to visualize several plots for right-censored data, as per example: the Kaplan-Meier estimate of the survival function, the smoothed estimate of the hazard function, the padronized residuals and forest plots associated to the coefficients of a Cox model, as well as the survival function estimates for parametric models. The vsd package has a modular design to allow adding more extensive function ality in the future. The plots created by the vsd package can be explored interactively, through an interactive mode with optional aid of the plotly package, or customized after creation by the user thanks to the usage of the ggplot2 graphical package, that allows for the easy re-utilization and customization of all created plots.
There’s a plethora of graphical packages on the R language ecosystem, specif ically in the area of Statistics and within the field of Survival Analysis, so much so that an user might not know where to start when considering a specific model or data structure. This dissertation details the functionality of a newly developed R package, vsd, meant to visualize data in Survival Analysis to ease this problem. It starts with an introduction on how to create R packages, with a focus on: building a package’s skeleton through the devtools package, in its docu mentation and consequent added functionality though the roxygen2 package, in some extra developmental concepts like linting for code correctness and package publication, and finally in how to use the version control system Git for easy availability of the package under an open license, as required by CRAN. Afterwards the discussion moves to a brief overview of any relevant con cepts under Survival Analysis for the developed package vsd. The vsd package, taking use of other packages on R’s ecosystem such as the packages survminer and flexsurv, allows the user to visualize several plots for right-censored data, as per example: the Kaplan-Meier estimate of the survival function, the smoothed estimate of the hazard function, the padronized residuals and forest plots associated to the coefficients of a Cox model, as well as the survival function estimates for parametric models. The vsd package has a modular design to allow adding more extensive function ality in the future. The plots created by the vsd package can be explored interactively, through an interactive mode with optional aid of the plotly package, or customized after creation by the user thanks to the usage of the ggplot2 graphical package, that allows for the easy re-utilization and customization of all created plots.
Description
Keywords
Análise de sobrevivência ggplot2 Pacotes do R R packages Survival analysis Matemática, Estatística e Aplicações . Faculdade de Ciências Exatas e da Engenharia