R/ex.R

In Marie-PerrotDockes/VariSel: Perform variable selection for different model types

# # require(tidyverse)
# # require(magrittr)
# # require(gganimate)
# # require(R6)
# # miaou <- QTLmod_group_univ$new(X,Y)
# # miaou$plot_coef()
# # miaou$plot_anime()
# # miaou$sel_cv()
# #
# #  miaou <- QTLmod_group_univ$new(X,Y)
# #  miaou$plot_cv()
# #  miaou <- QTLmod_group_multi_both$new(X,Y)
# #  miaou$plot_cv()
# #
# # miaou <- QTLmod_group_multi_marker$new(X,Y)
# #  miaou$plot_cv()
# #  miaou <- QTLmod_lasso$new(X,Y)
# #  miaou$plot_cv()
# #  miaou <-mod_lasso$new(X,Y,univ=FALSE)
#  # miaou$sel_cv()
# # types <- c("group_univ" ,
# # "group_multi_both",
# # "group_multi_marker",
# # "fused_univ",
# # "fused_multi_both",
# # "fused_multi_regr",
# # "lasso_univ",
# # "lasso_multi")
# # test <- lapply(types, function(type){
# #   print(type)
# #   chapeau(X,Y,type)
# # })
# #
# # group <- factor(c(rep("a", round(nrow(X)/2)),rep("b", nrow(X) - round(nrow(X)/2))))
# # # a<- chapeau(X,Y,"fus2mod_univ", group = group)
# # #
# require(rsample)
# require(tidyverse)
# iris2 <- iris %>% filter(Species %in% levels(Species)[1:2])
# rsamp <- iris2 %>%  select(-Species) %>%
#   bootstraps(5) %>% pull(splits) %>% first()
# resp <-c("Sepal.Length" ,"Sepal.Width")
#
# chapeau(iris2[,1:2], iris2[,3:4], type= "lasso_multi", resp, group = iris2$Species)
# #
#  X<-X[,700:800]
# # #
# ct <- compar_type(X, Y, types = c("fus2resp", "fus2mod_univ" , "group_univ" ,"group_multi_both" ,"group_multi_marker" ,
#                               "fused_univ" ,"fused_multi_both","fused_multi_regr",
#                               "lasso_univ","lasso_multi" ),group = iris2$Species[drop=TRUE])
#
# ### TO DO
# # Dans fus2resp il faut le mettre ne mode univ = FALSE , mais on lance le lasso en univ = TRE
# #on recupère en suite le modèle et on le retransforme
#
# # Créer une option "big" pour lasso et fus2 .
#
#  require (viridis)
# p <- ct %>% mutate(key = as.numeric(gsub('s','',key))) %>%
#   ggplot(aes(x = key, color = type,fill = type, y = value))+geom_smooth() + theme_bw() +
# labs(y = "MSE", title ="Bootstrap MSE", x = "Regularization Path") + facet_grid(~Trait) +
#   scale_color_viridis( discrete = TRUE) +scale_fill_viridis( discrete = TRUE)
# ggsave(p , file = "ex_bootmse.pdf")
# group <- iris$Species
# X <- as.matrix(iris[,1:2])
# X1  <- model.matrix(~group + group:X - 1)
# X1 %>% as.tibble() %>% gather() %>%
#   mutate(Group = str_extract_all(key,paste(grp_value,collapse="|"))) %>% pull(Group)
#
# paste0('{',paste(paste0("group",unique(group)), collapse =","),'}')
#
# plan(sequential) ;
# ct <- compar_type(X[,1:10], Y, types = c("fus2resp",  "group_univ" ,"group_multi_both" ,"group_multi_marker" ,
#                                          "fused_univ" ,"fused_multi_both","fused_multi_regr",
#                                          "lasso_univ","lasso_multi" ))
#
# lapin <- function(a){
# u <- a +4
#     browser()
#
#   print('mieou')
# }