Step_dummy. Dealing with duplicated column names generated by recipe() steps, Tidymodels - r

Question

Dear community,
I have been struggeling for extensive amount of time now trying to understand what is going on here, when I perform my recipe() steps for my linear (glm) model using the Tidymodels framework. The recipe() step_dummy(all_nominal(), -all_outcomes()) was suggested by the usemodels() function https://usemodels.tidymodels.org/index.html .
When I commend out the step_dummy() the recipe() and prep() works fine, however its important to me that these categorical variables are dummyfied (..is that a word!?).
This is the first time I making use of and including a reprex in a question on stackoverflow, so please let me know if you need more information to assist on this matter.
I have looked everywhere, e.g. including a one_hot = TRUE or keep_original_cols argument in the step_dummy() but it does not seem to be effective.
It should be quite easy as it is a matter of renaming the generated columns as unique, but do not succeed. Here is the era.af_train set.
> era.af_train
# A tibble: 7,104 x 44
logRR ID AEZ16simple PrName.Code SubPrName.Code Product Country
<dbl> <dbl> <fct> <fct> <fct> <fct> <fct>
1 -0.851 1663 Warm.Semiar~ BP TW Pearl Mill~ Niger
2 -1.17 1663 Warm.Semiar~ BP/Mu Mu-N/TW Pearl Mill~ Niger
3 -0.314 1663 Warm.Semiar~ BP TW Pearl Mill~ Niger
4 -0.776 1663 Warm.Semiar~ BP TW Pearl Mill~ Niger
5 -0.0850 1675 Warm.Semiar~ AP TPM+N Pearl Mill~ Niger
6 -0.159 1689 Warm.Subhum~ Al/AP/BP Al+N/LF/TP/TPM~ Maize Togo
7 -0.579 1701 Warm.Semiar~ BP TW Fodder (Le~ Tunisia
8 -0.662 1729 Warm.Subhum~ Al Al-N/Al+N Cassava or~ Nigeria
9 -1.80 1802 Cool.Subhum~ Al/AP Al+N/TP Wheat Ethiop~
10 -1.74 1802 Cool.Subhum~ Al/AP Al+N/TP/TPI+N Wheat Ethiop~
# ... with 7,094 more rows, and 37 more variables: Latitude <dbl>,
# Longitude <dbl>, Site.Type <fct>, Tree <fct>, Bio01_MT_Anu.Mean <dbl>,
# Bio02_MDR.Mean <dbl>, Bio03_Iso.Mean <dbl>, Bio04_TS.Mean <dbl>,
# Bio05_TWM.Mean <dbl>, Bio06_MinTCM.Mean <dbl>, Bio07_TAR.Mean <dbl>,
# Bio08_MT_WetQ.Mean <dbl>, Bio09_MT_DryQ.Mean <dbl>,
# Bio10_MT_WarQ.Mean <dbl>, Bio11_MT_ColQ.Mean <dbl>,
# Bio12_Pecip_Anu.Mean <dbl>, Bio13_Precip_WetM.Mean <dbl>,
# Bio14_Precip_DryM.Mean <dbl>, Bio15_Precip_S.Mean <dbl>,
# Bio16_Precip_WetQ.Mean <dbl>, Bio17_Precip_DryQ.Mean <dbl>,
# Mean_log.n_tot_ncs <dbl>, Mean_log.ca_mehlich3 <dbl>,
# Mean_log.k_mehlich3 <dbl>, Mean_log.mg_mehlich3 <dbl>,
# Mean_log.p_mehlich3 <dbl>, Mean_log.s_mehlich3 <dbl>,
# Mean_log.fe_mehlich3 <dbl>, Mean_db_od <dbl>, Mean_bdr <dbl>,
# Mean_sand_tot_psa <dbl>, Mean_clay_tot_psa <dbl>, Mean_ph_h2o <dbl>,
# Mean_log.ecec.f <dbl>, Mean_log.c_tot <dbl>, Mean_log.oc <dbl>,
# Slope.mean <dbl>
I am including the columns ID, AEZ16simple, PrName.Code, SubPrName.Code, Product, Country, Latitude and Longitude as "ID variables", as I wish to compare the glm model later with a random forest model and a XGBoost model.
All help is welcome!
Have a good weekend and
thank you in advance.
library(reprex)
#> Warning: package 'reprex' was built under R version 4.0.5
library(dplyr)
#> Warning: package 'dplyr' was built under R version 4.0.5
#>
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#>
#> filter, lag
#> The following objects are masked from 'package:base':
#>
#> intersect, setdiff, setequal, union
library(recipes)
#> Warning: package 'recipes' was built under R version 4.0.5
#>
#> Attaching package: 'recipes'
#> The following object is masked from 'package:stats':
#>
#> step
library(readr)
#> Warning: package 'readr' was built under R version 4.0.5
setwd("C:/Users/lindh011/OneDrive - Wageningen University & Research/Internship ICRAF (ERA)/ERA_Agroforestry_WURwork")
era.af_train <- read_csv("era.af_train.csv")
#>
#> -- Column specification --------------------------------------------------------
#> cols(
#> .default = col_double(),
#> AEZ16simple = col_character(),
#> PrName.Code = col_character(),
#> SubPrName.Code = col_character(),
#> Product = col_character(),
#> Country = col_character(),
#> Site.Type = col_character(),
#> Tree = col_character()
#> )
#> i Use `spec()` for the full column specifications.
era.af_train_Tib <- as_tibble(era.af_train)
glmnet_recipe <-
recipe(formula = logRR ~ ., data = era.af_train) %>%
step_novel(all_nominal(), -all_outcomes()) %>%
step_dummy(all_nominal(), -all_outcomes(), naming = dummy_names) %>%
step_zv(all_predictors()) %>%
step_normalize(all_predictors(), -all_nominal()) %>%
update_role(ID,
AEZ16simple,
PrName.Code,
SubPrName.Code,
Product,
Country,
Latitude,
Longitude,
new_role = "sample ID") %>%
step_impute_mode(all_nominal(), -all_outcomes()) %>%
step_impute_knn (all_numeric_predictors()) %>%
step_impute_knn(logRR) %>%
step_corr(all_numeric_predictors()) %>%
step_nzv(all_numeric_predictors()) %>%
prep()
#> Error: Column names `SubPrName.Code_AF.N.Al.N.TP`, `SubPrName.Code_AF.N.Al.N.TP.TPM`, `SubPrName.Code_Al.N.In.N`, `SubPrName.Code_Al.N.In.N`, `SubPrName.Code_Al.N`, and 33 more must not be duplicated.
#> Use .name_repair to specify repair.
Created on 2021-07-02 by the reprex package (v2.0.0)