Global and Local Explanation Functions Do Not Work With Datasets Have Any Missing Values

[aalhadab]

When a dataset has some any missing values, global and local functions don't work because the generated prediction matrix has different dimensions. This problem occurs even when na.action = "na.impute" is used.

library(mlr3verse)
#> Warning: package 'mlr3verse' was built under R version 4.1.2
#> Loading required package: mlr3
library(tidyverse)
#> Warning: package 'tidyverse' was built under R version 4.1.2
#> Warning: package 'ggplot2' was built under R version 4.1.2
#> Warning: package 'tibble' was built under R version 4.1.2
#> Warning: package 'tidyr' was built under R version 4.1.2
#> Warning: package 'readr' was built under R version 4.1.2
#> Warning: package 'purrr' was built under R version 4.1.2
#> Warning: package 'dplyr' was built under R version 4.1.2
#> Warning: package 'stringr' was built under R version 4.1.2
#> Warning: package 'forcats' was built under R version 4.1.2
library(randomForestSRC)
#> 
#>  randomForestSRC 3.0.2 
#>  
#>  Type rfsrc.news() to see new features, changes, and bug fixes. 
#> 
#> 
#> Attaching package: 'randomForestSRC'
#> The following object is masked from 'package:purrr':
#> 
#>     partial
#> The following object is masked from 'package:mlr3verse':
#> 
#>     tune
library(survival)
library(survxai)
#> Loading required package: prodlim
#> Warning: package 'prodlim' was built under R version 4.1.2
#> Welcome to survxai (version: 0.2.1).
#> Information about the package can be found in the GitHub repository: https://github.com/MI2DataLab/survxai
#> 
#> Attaching package: 'survxai'
#> The following object is masked from 'package:dplyr':
#> 
#>     explain

task  = tsk("actg")
task2 = tsk("lung")

os   = rfsrc(Surv(time, status)~. , task$data(), na.action = "na.impute", importance = "permute") 

os2  = rfsrc(Surv(time, status)~. , task2$data(), na.action = "na.impute", importance = "permute")

rsf_explain = explain(model = os, 
                     data  = task$data()[, -c(1:2)],
                     y     = Surv(task$data()$time, task$data()$status))

rsf_explain2= explain(model = os2, 
                     data  = task2$data()[, -c(1:2)],
                     y     = Surv(task2$data()$time, task2$data()$status))

model_performance(rsf_explain)  %>% plot()
#> No covariates  specified: Kaplan-Meier for censoring times used for weighting.

model_performance(rsf_explain2) %>% plot()
#> No covariates  specified: Kaplan-Meier for censoring times used for weighting.
#> No covariates  specified: Kaplan-Meier for censoring times used for weighting.
#> Error in predictSurvProb.rfsrc(X.model, newdata, times): 
#> Prediction matrix has wrong dimensions:
#> Requested newdata x times: 228 x 187
#> Provided prediction matrix: 167 x 187

^{Created on 2022-03-28 by the reprex package (v2.0.1)}

[hbaniecki]

Hi @aalhadab, you can check out our new updated package at https://github.com/ModelOriented/survex.

[agosiewska]

As far as I am aware, the deprecation of the survxai package is not the official statement of the package maintainer @AleksandraDabrowska . Good manners and programming practices would require at least asking the author.