I am either too caffeinated or not caffeinated enough, because I cannot figure out how to do this. I need to create a function that calculates an equation that exponentiates the intercept and effect for several sets of variables, each set grouped by a common string within column names, and then sums all the exponents, yielding a single value. I need to do this across columns within each row so dplyr seems the obvious choice. The tricky part is that the function needs to be able to do this for a different number of elements within each set. Easier to show than describe.
Here are two datasets
set.seed(1)
names_df1 <- c("ball", "bell", "bat")
df1 <- data.frame(int_ball = sample(seq(-.99,-.01, .01),5,replace=T),
eff_ball = sample(seq(-.99,-.01, .01),5,replace=T),
int_bell = sample(seq(-.99,-.01, .01),5,replace=T),
eff_bell = sample(seq(-.99,-.01, .01),5,replace=T),
int_bat = sample(seq(-.99,-.01, .01),5,replace=T),
eff_bat = sample(seq(-.99,-.01, .01),5,replace=T))
names_df2 <- c("dog", "cat", "bird", "fish")
df2 <- data.frame(int_dog = sample(seq(-.99,-.01, .01),5,replace=T),
eff_dog = sample(seq(-.99,-.01, .01),5,replace=T),
int_cat = sample(seq(-.99,-.01, .01),5,replace=T),
eff_cat = sample(seq(-.99,-.01, .01),5,replace=T),
int_bird = sample(seq(-.99,-.01, .01),5,replace=T),
eff_bird = sample(seq(-.99,-.01, .01),5,replace=T),
int_fish = sample(seq(-.99,-.01, .01),5,replace=T),
eff_fish = sample(seq(-.99,-.01, .01),5,replace=T))
Each dataset has as many pairs of variables as there are elements in the string vector preceding each dataset (names_df1 and names_df2). I need to add together the int_ and eff_ variables for each pair, then exponentiate the result, then add all those exponents together.
For the dataset we three sets of pairs the results would look like this
df1 %>%
mutate(eq_df1 = exp(int_ball + eff_ball) + exp(int_bell + eff_bell) + exp(int_bat + eff_bat))
# int_ball eff_ball int_bell eff_bell int_bat eff_bat eq_df1
# 1 -0.32 -0.57 -0.03 -0.93 -0.11 -0.21 1.519698
# 2 -0.61 -0.86 -0.15 -0.27 -0.63 -0.67 1.159504
# 3 -0.99 -0.18 -0.79 -0.21 -0.66 -0.16 1.118678
# 4 -0.66 -0.41 -0.46 -0.15 -0.11 -0.65 1.354026
# 5 -0.13 -0.49 -0.26 -0.63 -0.56 -0.30 1.371762
And for the dataset with four sets of pairs it would look like this
df2 %>%
mutate(eq_df2 = exp(int_dog + eff_dog) + exp(int_cat + eff_cat) + exp(int_bird + eff_bird) + exp(int_fish + eff_fish))
# int_dog eff_dog int_cat eff_cat int_bird eff_bird int_fish eff_fish eq_df2
# 1 -0.26 -0.80 -0.56 -0.58 -0.98 -0.35 -0.19 -0.11 1.671570
# 2 -0.58 -0.56 -0.75 -0.94 -0.55 -0.30 -0.87 -0.77 1.125734
# 3 -0.62 -0.13 -0.30 -0.76 -0.82 -0.13 -0.60 -0.16 1.673230
# 4 -0.80 -0.30 -0.61 -0.68 -0.78 -0.30 -0.11 -0.71 1.388169
# 5 -0.72 -0.60 -0.49 -0.86 -0.22 -0.25 -0.52 -0.87 1.400453
Any help much appreciated. The solution doesn't have to be in dplyr.
You could define your function to pivots the cols to long format, do the required calculation, and bind back to the original data:
library(dplyr)
library(tidyr)
library(tibble)
f <- function(.data, vars = starts_with(c("eff_", "int_"))) {
.data |>
select( {{ vars }} ) |>
rowid_to_column() |>
pivot_longer(-rowid, names_sep = "_", names_to = c(".value", "name")) |>
summarise(eq = sum(exp(pick(2) + pick(3))), .by = rowid) |>
select(-rowid) |>
bind_cols(.data, results = _)
}
f(df1)
int_ball eff_ball int_bell eff_bell int_bat eff_bat eq
1 -0.32 -0.57 -0.03 -0.93 -0.11 -0.21 1.519698
2 -0.61 -0.86 -0.15 -0.27 -0.63 -0.67 1.159504
3 -0.99 -0.18 -0.79 -0.21 -0.66 -0.16 1.118678
4 -0.66 -0.41 -0.46 -0.15 -0.11 -0.65 1.354026
5 -0.13 -0.49 -0.26 -0.63 -0.56 -0.30 1.371762
f(df2)
int_dog eff_dog int_cat eff_cat int_bird eff_bird int_fish eff_fish eq
1 -0.26 -0.80 -0.56 -0.58 -0.98 -0.35 -0.19 -0.11 1.671570
2 -0.58 -0.56 -0.75 -0.94 -0.55 -0.30 -0.87 -0.77 1.125734
3 -0.62 -0.13 -0.30 -0.76 -0.82 -0.13 -0.60 -0.16 1.673230
4 -0.80 -0.30 -0.61 -0.68 -0.78 -0.30 -0.11 -0.71 1.388169
5 -0.72 -0.60 -0.49 -0.86 -0.22 -0.25 -0.52 -0.87 1.400453