Reputation: 772
Loop with if else and next R
I am combining a loop with if else and next in R.
In order to reproduce my problem in its complexity, I cannot provide a minimum example but a big piece of code. The aim is to fill the list df with min, max, 50percentile in the columns rmse_1 and rmse_2.
You need to change the path according to your desired path at the positions marked with # !!! change path. If you changed the paths you can run the code:
# create lists
mse_samp <- list("mse_A" = list("P10" = data.frame(number = seq(1,3,1),
mse_1 = c(2.5, 4.6, 7.8),
mse_2 = c(6.7, 8.9, 4.1)),
"P30" = data.frame(number = seq(1,3,1),
mse_1 = c(22.5, 74.6, 97.8),
mse_2 = c(56.7, 78.9, 14.1))),
"mse_B" = list("P10" = data.frame(number = seq(1,3,1),
mse_1 = c(122.5, 124.6, 127.8),
mse_2 = c(126.7, 128.9, 124.1)),
"P30" = data.frame(number = seq(1,3,1),
mse_1 = c(3422.5, 3474.6, 3497.8),
mse_2 = c(3456.7, 3478.9, 3414.1))))
# !!! change path
save(mse_samp,
file="H:\\R\\Forum_data\\dat1.RData")
mse_samp <- list("mse_A" = list("P70" = data.frame(number = seq(1,3,1),
mse_1 = c(22.2, 77.6, 97.8, 21.2, 73.9),
mse_2 = c(26.7, 78.9, 17.1, 23.2, 82.2)),
"P80" = data.frame(number = seq(1,3,1),
mse_1 = c(1022.2, 3077.6, 9097.8, 1221.2, 7373.9),
mse_2 = c(7626.7, 2278.9, 7317.1, 7623.2, 8982.2))),
"mse_B" = list("P70" = data.frame(number = seq(1,3,1),
mse_1 = c(3722.2, 3777.6, 3797.8),
mse_2 = c(3726.7, 3778.9, 3717.1)),
"P80" = data.frame(number = seq(1,3,1),
mse_1 = c(1022.2, 3077.6, 9097.8),
mse_2 = c(7626.7, 2278.9, 7317.1))))
save(mse_samp,
file="H:\\R\\Forum_data\\dat2.RData")
# create table for min max for different perc and runs for each paramter (loop)
n_measure <- 3 # number of different measures
npr1 <- 2 # number of different percs run1
npr2 <- 2 # number of different percs run2
targets <- c("A", "B")
for (i in 1:length(targets)) {
df <- data.frame(run = c(rep("run1", n_measure * npr1),
rep("run2", n_measure * npr2)),
perc_train = c(rep(c(0.1, 0.3), times = 1, each = n_measure), # percs run 1
rep(c(0.7, 0.8), times = 1, each = n_measure)), # percs run 2
measure = c(rep(c("min", "max", "50percentile"),
times = npr1 + npr2, each = 1)),
rmse_1 = rep(NA, n_measure * (npr1 + npr2)),
rmse_2 = rep(NA, n_measure * (npr1 + npr2))
)
assign(paste0('df_', targets[i]), df)
}
df <- list("A" = df_A, "B" = df_B)
# convert column which are factors to characters
for (i in 1:length(targets)) {
df[[i]][sapply(df[[i]], is.factor)] <- lapply(df[[i]][sapply(df[[i]], is.factor)],
as.character)
}
rm(list = c("df_A", "df_B", "df_C"))
# !!! change path
path <- c("H:\\R\\Forum_data\\dat1.RData", # run1
# !!! change path
"H:\\R\\Forum_data\\dat2.RData") # run2
percs_names <- c("P10", "P30", "P70", "P80")
percs <- c(0.1, 0.3, 0.7, 0.8)
targets <- c("A", "B")
run_name <- c("run1", "run2")
measure_name <- c("min", "max", "50percentile")
fill_names <- c("rmse_min_1", "rmse_min_2", "rmse_max_1", "rmse_max_2",
"percentile_50_1", "percentile_50_2")
var_name <- c("rmse_1", "rmse_2")
a_or_b <- c("a","b")
# read in data
for (i in 1:length(path)) {
load(path[i])
dat <- mse_samp
for (j in 1:length(targets)) {
for (k in 1:length(percs_names)) {
# if statement
if(percs_names[k] == names(dat[[j]][k])){
dat1 <- dat[[paste0("mse_", targets[j])]][k][[1]]
rmse_min_1 <- sqrt(min(dat1$mse_1))
rmse_min_2 <- sqrt(min(dat1$mse_2))
rmse_max_1 <- sqrt(max(dat1$mse_1))
rmse_max_2 <- sqrt(max(dat1$mse_2))
percentile_50_1 <- quantile(sqrt(dat1$mse_1), probs = 0.5)
percentile_50_2 <- quantile(sqrt(dat1$mse_2), probs = 0.5)
for (fi in 1:length(fill_names)) {
for (m in 1:length(measure_name)) {
a <- which(df[[targets[j]]]$run == run_name[i] &
df[[targets[j]]]$measure == measure_name[m] &
df[[targets[j]]]$perc_train == percs[k] &
is.na(df[[targets[j]]]$rmse_1)
)
b <- which(df[[targets[j]]]$run == run_name[i] &
df[[targets[j]]]$measure == measure_name[m] &
df[[targets[j]]]$perc_train == percs[k] &
is.na(df[[targets[j]]]$rmse_2)
)
for (v in 1:length(var_name)) {
df[[targets[j]]][eval(parse(text = a_or_b[v])), which(names(df[[targets[j]]]) == var_name[v])] <- eval(parse(text = fill_names[fi]))
}
}
}
}
else { next }
}
}
}
1. issue After running the code, the following error message occurs:
Error in if (percs_names[k] == names(dat[[j]][k])) { :
missing value where TRUE/FALSE needed
I guess the error is probably caused by the if elsestatement. How can I run the code without the error?
2. issue Currently only rows for run1 are filled. The rmse_1 an rmse_2 are filled with the same values in rows min, max, 50percentile. They should be different. How can I fill the other runs and fill the rows correctly? In the end there should be no NA left.
Upvotes: 0
Views: 1203
Answers (1)
Reputation: 7724
Although you insist on using for-loops here is a solution to you problem with map (similiar to lapply) and some tidyverse-magic.
I have one assumption: All datasets your working on are stored in a list, called data_runs_list. An example is given at the end of the answer in the Data-section (using your example data).
So lets first bring this nested structure in a more readable format:
library(tidyverse)
library(stringr)
data_runs_df <-
map(data_runs_list, ~ map(.x, bind_rows, .id = "perc") %>%
bind_rows(.id = "target")) %>%
bind_rows(.id = "run")
data_runs_df
# A tibble: 24 x 6
# run target perc number mse_1 mse_2
# <chr> <chr> <chr> <int> <dbl> <dbl>
# 1 run1 mse_A P10 1 2.5 6.7
# 2 run1 mse_A P10 2 4.6 8.9
# 3 run1 mse_A P10 3 7.8 4.1
# 4 run1 mse_A P30 1 22.5 56.7
# 5 run1 mse_A P30 2 74.6 78.9
# 6 run1 mse_A P30 3 97.8 14.1
# 7 run1 mse_B P10 1 122. 127.
# 8 run1 mse_B P10 2 125. 129.
# 9 run1 mse_B P10 3 128. 124.
# 10 run1 mse_B P30 1 3422. 3457.
# # ... with 14 more rows
To better understand, what bind_rows() does, only take the first entry of the first entry of list and see what happens:
bind_rows(data_runs_list[[1]][[1]], .id = "perc")
# perc number mse_1 mse_2
# 1 P10 1 2.5 6.7
# 2 P10 2 4.6 8.9
# 3 P10 3 7.8 4.1
# 4 P30 1 22.5 56.7
# 5 P30 2 74.6 78.9
# 6 P30 3 97.8 14.1
The two dataframes are stacked together and the id-column perc preserves the original list name. map then applies to each level of the list bind_row one after another, on each level with a different id-column.
So this is something to work with nicely. You want to have min, max and 50%-quantile (i.e. the median) of for each run, percentage, target and seperately for the two measurements mse_1 and mse_2. group_by is perfect for this in combindation with summarize. To better handle the two different measurements, first convert the data to a long format. If you have more measurements you can just specify them at the end of the gather-call:
data_runs_df <- data_runs_df %>%
gather(mse, value, mse_1, mse_2)
data_runs_df
# A tibble: 48 x 6
# run target perc number mse value
# <chr> <chr> <chr> <int> <chr> <dbl>
# 1 run1 mse_A P10 1 mse_1 2.5
# 2 run1 mse_A P10 2 mse_1 4.6
# 3 run1 mse_A P10 3 mse_1 7.8
# 4 run1 mse_A P30 1 mse_1 22.5
# 5 run1 mse_A P30 2 mse_1 74.6
# 6 run1 mse_A P30 3 mse_1 97.8
# 7 run1 mse_B P10 1 mse_1 122.
# 8 run1 mse_B P10 2 mse_1 125.
# 9 run1 mse_B P10 3 mse_1 128.
# 10 run1 mse_B P30 1 mse_1 3422.
# ... with 38 more rows
Now, before the measurements are calculated, we rename the target and mse column quick and then use the group_by in combindation with summarize:
data_info <- data_runs_df %>%
mutate(mse = str_c("r", mse),
target = str_remove(target, "mse_")) %>%
group_by(run, target, perc, mse) %>%
summarize(min = min(sqrt(value)),
max = max(sqrt(value)),
median = median(sqrt(value)))
data_info
# A tibble: 16 x 7
# Groups: run, target, perc [?]
# run target perc mse min max median
# <chr> <chr> <chr> <chr> <dbl> <dbl> <dbl>
# 1 run1 A P10 rmse_1 1.58 2.79 2.14
# 2 run1 A P10 rmse_2 2.02 2.98 2.59
# 3 run1 A P30 rmse_1 4.74 9.89 8.64
# 4 run1 A P30 rmse_2 3.75 8.88 7.53
# 5 run1 B P10 rmse_1 11.1 11.3 11.2
# 6 run1 B P10 rmse_2 11.1 11.4 11.3
# 7 run1 B P30 rmse_1 58.5 59.1 58.9
# 8 run1 B P30 rmse_2 58.4 59.0 58.8
# 9 run2 A P70 rmse_1 4.71 9.89 8.81
# 10 run2 A P70 rmse_2 4.14 8.88 5.17
# 11 run2 A P80 rmse_1 32.0 95.4 55.5
# 12 run2 A P80 rmse_2 47.7 87.3 85.5
# 13 run2 B P70 rmse_1 61.0 61.6 61.5
# 14 run2 B P70 rmse_2 61.0 61.5 61.0
# 15 run2 B P80 rmse_1 32.0 95.4 55.5
# 16 run2 B P80 rmse_2 47.7 87.3 85.5
Now one last step to get everything it the exact shap you wanted to have it we need gather and its counterpart spread:
data_info <- data_info %>%
gather(measure, value, min, max, median) %>%
spread(mse, value)
data_info
# A tibble: 24 x 6
# Groups: run, target, perc [8]
# run target perc measure rmse_1 rmse_2
# <chr> <chr> <chr> <chr> <dbl> <dbl>
# 1 run1 A P10 max 2.79 2.98
# 2 run1 A P10 median 2.14 2.59
# 3 run1 A P10 min 1.58 2.02
# 4 run1 A P30 max 9.89 8.88
# 5 run1 A P30 median 8.64 7.53
# 6 run1 A P30 min 4.74 3.75
# 7 run1 B P10 max 11.3 11.4
# 8 run1 B P10 median 11.2 11.3
# 9 run1 B P10 min 11.1 11.1
# 10 run1 B P30 max 59.1 59.0
# ... with 14 more rows
Everyting in two calls:
data_runs_df <-
map(data_runs_list, ~ map(.x, bind_rows, .id = "perc") %>%
bind_rows(.id = "target")) %>%
bind_rows(.id = "run")
data_info <- data_runs_df %>%
gather(mse, value, mse_1, mse_2) %>%
mutate(mse = str_c("r", mse),
target = str_remove(target, "mse_")) %>%
group_by(run, target, perc, mse) %>%
summarize(min = min(sqrt(value)),
max = max(sqrt(value)),
median = median(sqrt(value))) %>%
gather(measure, value, min, max, median) %>%
spread(mse, value)
If you insist on the list-format you were using, you can do:
data_info_list <- map(c("A", "B"), function(x) filter(data_info, target == x))
names(data_info_list) <- c("A", "B")
Data
mse_samp1 <-
list("mse_A" = list("P10" = data.frame(number = 1:3, mse_1 = c(2.5, 4.6, 7.8), mse_2 = c(6.7, 8.9, 4.1)),
"P30" = data.frame(number = 1:3, mse_1 = c(22.5, 74.6, 97.8), mse_2 = c(56.7, 78.9, 14.1))),
"mse_B" = list("P10" = data.frame(number = 1:3, mse_1 = c(122.5, 124.6, 127.8), mse_2 = c(126.7, 128.9, 124.1)),
"P30" = data.frame(number = 1:3, mse_1 = c(3422.5, 3474.6, 3497.8), mse_2 = c(3456.7, 3478.9, 3414.1))))
mse_samp2 <-
list("mse_A" = list("P70" = data.frame(number = 1:3, mse_1 = c(22.2, 77.6, 97.8), mse_2 = c(26.7, 78.9, 17.1)),
"P80" = data.frame(number = 1:3, mse_1 = c(1022.2, 3077.6, 9097.8), mse_2 = c(7626.7, 2278.9, 7317.1))),
"mse_B" = list("P70" = data.frame(number = 1:3, mse_1 = c(3722.2, 3777.6, 3797.8), mse_2 = c(3726.7, 3778.9, 3717.1)),
"P80" = data.frame(number = 1:3, mse_1 = c(1022.2, 3077.6, 9097.8), mse_2 = c(7626.7, 2278.9, 7317.1))))
data_runs_list <- list(run1 = mse_samp1, run2 = mse_samp2)
Upvotes: 3
Related Questions
- R An if else statement inside a for loop
- Nested for-loop with if statement
- How to do nested for loops with an if
- Nested loop if in r
- How to Implement a Complex For-Loop + If Statement
- if else inside for loop
- Nested For and If Loops in R
- Nested For Loops w/ If-Else
- Nested if-else loops in R
- For loop with conditional if statements