Testing imputation methods on GDC TCGA clinical data

tcga_impute_sandbox.R

library(missForest)
library(mice)
library(ggplot2)
library(ggmice)

input_df <- gdc_case_meta[
    c("project_id", "gender", "age_at_diagnosis", "tumor_stage")
]

input_df$project_id <- factor(input_df$project_id)
input_df$gender <- factor(input_df$gender)
input_df$tumor_stage <- ordered(
    input_df$tumor_stage,
    levels = c("i", "ii", "iii", "iv")
)

mf <- missForest(input_df, maxiter = 100, verbose = TRUE)
imp_df <- mf$ximp

me <- mice(input_df, m = 1)

ggmice(me, aes(age_at_diagnosis, project_id)) + geom_point()