how to fix undefined column selected error in R?

Question

I am not quite use caret r package for my project though, I intend to make prediction using either lasso or randomforest. I used randomforest to do prediction on my data, but I got weird error as follow:

>     Error in `[.data.frame`(data, , all.vars(Terms), drop = FALSE) : 
>       undefined columns selected
>     In addition: There were 50 or more warnings (use warnings() to see the first 50)

I don't understand why it is happening. any clue to make this work? why I have this error? any thought?

minimal reproducible data

here is the minimal reproducible data:

mydf = structure(list(taken_time = c(15L, 5L, 39L, 
-21L, 46L, 121L, 9L, 100L, 70L, 92L, 31L, 37L), ap6xl = c(203.2893857, 
4.858269406, 200, 14220, 218.2215352, 115.5227706, 4.858269406, 
516.18125, 72.06166523, 4.858269406, 96.68516046, 386.1480917
), pct5 = c(732.074484, 25.67901235, 1900, 120.0477168, 3621.328567, 
79.30561111, 8376.70314, 4183.709089, 59.77649029, 997.7490228, 
118.9774144, 171.2285804), crp4 = c(196115424.7, 1073624.455, 
10007, 1457496.474, 10343851.7, 81288042.73, 320405225.1, 334807893.9, 
112950094.2, 15775090.31, 3008739.881, 127837638.1), age = c(52L, 
74L, 52L, 67L, 82L, 67L, 71L, 84L, 58L, 52L, 81L, 60L), gender = structure(c(2L, 
2L, 2L, 1L, 2L, 1L, 1L, 1L, 1L, 2L, 2L, 2L), .Label = c("F", 
"M"), class = "factor"), inpatient_readmission_time_rtd = c(79.78819444, 
57.59068053, 57.59068053, 57.59068053, 57.59068053, 9.893055556, 
150.1951389, 57.59068053, 134.05625, 57.59068053, 65.16041667, 
17.46527778), infection_flag = c(0L, 0L, 1L, 1L, 0L, 1L, 0L, 
1L, 1L, 1L, 1L, 0L), temperature_value = c(98.9, 98.9, 98, 101.3, 
99.5, 98.1, 98.7, 97.1, 98.1, 98.2, 100.4, 98.8), heartrate_value = c(106, 
61, 78, 91, 120, 68, 93.55081001, 122, 110, 75, 116, 111), pH_result_time_rta = c(11, 
85.50402145, 85.50402145, 85.50402145, 85.50402145, 85.50402145, 
85.50402145, 85.50402145, 85.50402145, 85.50402145, 50, 85.50402145
), gcst_value = c(15, 15, 15, 14.63769293, 15, 14.63769293, 15, 
15, 15, 14.63769293, 15, 15)), row.names = c(NA, 12L), class = "data.frame")

my attempt

this is what I tried, but caret is simply complaining about that. why? any idea?

library(caret)

fitControl <- trainControl(method = "repeatedcv",number = 10,repeats = 10, search = "random")
model_cv <- train(mydf$gcst_value ~ .,data = dat,method = "randomforest",
                  trControl = fitControl,na.action = na.omit)

immunoscore = predict(model_cv, mydf)

update:

here is my r session:

> > sessionInfo() R version 3.6.3 (2020-02-29) Platform: x86_64-w64-mingw32/x64 (64-bit) Running under: Windows 10 x64 (build
> 18362)
> 
> Matrix products: default
> 
> Random number generation:  RNG:     Mersenne-Twister   Normal: 
> Inversion   Sample:  Rounding    locale: [1] LC_COLLATE=English_United
> States.1252  LC_CTYPE=English_United States.1252    [3]
> LC_MONETARY=English_United States.1252 LC_NUMERIC=C                   
> [5] LC_TIME=English_United States.1252    
> 
> attached base packages: [1] stats     graphics  grDevices utils    
> datasets  methods   base     
> 
> other attached packages: [1] randomForest_4.6-14 data.table_1.12.8  
> stringr_1.4.0       ranger_0.12.1       caret_6.0-86        [6]
> ggplot2_3.3.0       lattice_0.20-38     jsonlite_1.6.1     
> dplyr_0.8.5        
> 
> loaded via a namespace (and not attached):  [1] Rcpp_1.0.3          
> pillar_1.4.3         compiler_3.6.3       gower_0.2.1         
> plyr_1.8.6            [6] class_7.3-15         iterators_1.0.12    
> tools_3.6.3          elasticnet_1.1.1     rpart_4.1-15         [11]
> ipred_0.9-9          lubridate_1.7.4      lifecycle_0.2.0     
> tibble_2.1.3         gtable_0.3.0         [16] nlme_3.1-144        
> pkgconfig_2.0.3      rlang_0.4.5          Matrix_1.2-18       
> foreach_1.5.0        [21] rstudioapi_0.11      prodlim_2019.11.13  
> withr_2.1.2          pROC_1.16.2          generics_0.0.2       [26]
> recipes_0.1.10       stats4_3.6.3         nnet_7.3-12         
> grid_3.6.3           tidyselect_1.0.0     [31] glue_1.3.2          
> R6_2.4.1             survival_3.1-8       lava_1.6.7          
> reshape2_1.4.3       [36] purrr_0.3.3          magrittr_1.5        
> lars_1.2             ModelMetrics_1.2.2.2 splines_3.6.3        [41]
> MASS_7.3-51.5        scales_1.1.0         codetools_0.2-16    
> assertthat_0.2.1     timeDate_3043.102    [46] colorspace_1.4-1    
> stringi_1.4.6        munsell_0.5.0        crayon_1.3.4

Answer 1

You need to fix two issues:

• You need to have all columns in data. This causes the error in your question since gcst_value is in a different data.frame than the data argument(dat)

• randomForest is not a valid model. It is represented by rf in the method argument.

Fixing the above issues(see notes below):

fitControl <- trainControl(method = "repeatedcv",number = 10,repeats = 10, 
   search = "random")
    model_cv <- train(gcst_value ~ .,data = mydf,method = "rf",
                      trControl = fitControl,
    na.action = na.omit)
    immunoscore = predict(model_cv, mydf)

Summary:

 summary(model_cv)
                Length Class      Mode     
call              4    -none-     call     
type              1    -none-     character
predicted        12    -none-     numeric  
mse             500    -none-     numeric  
rsq             500    -none-     numeric  
oob.times        12    -none-     numeric  
importance       11    -none-     numeric  
importanceSD      0    -none-     NULL     
localImportance   0    -none-     NULL     
proximity         0    -none-     NULL  

Getting the RMSE(purely representative)

RMSE(immunoscore,mydf$gcst_value)
[1] 0.08737056

NOTE

1. The validity of this model is the responsibility of the original poster.

2. Warnings are likely due to issues with model validity. I omitted those from the answer.

Further Note

On inspecting the warning messages(see note 1 above):

50: In randomForest.default(x, y, mtry = param$mtry, ...) : The response has five or fewer unique values. Are you sure you want to do regression?