在 function 中按组使用 lapply，在 R 中使用 2 arguments

Question

I would like to calculate the roc of several variables in different groups, fixing only the response variable, here's what I've been trying below:

library(pROC)
data(aSAH)
lapply(dplyr::select(aSAH,c(s100b,ndka)),roc,response = aSAH$outcome)

And output:

$s100b

Call:
roc.default(response = ..1, predictor = X[[i]])

Data: X[[i]] in 72 controls (..1 Good) < 41 cases (..1 Poor).
Area under the curve: 0.7314

$ndka

Call:
roc.default(response = ..1, predictor = X[[i]])

Data: X[[i]] in 72 controls (..1 Good) < 41 cases (..1 Poor).
Area under the curve: 0.612

But i need to apply this in each gender and in selected variables. Just like a group_by(gender) %>% roc()

Thanks!

Answer 1

Using by .

by(aSAH, aSAH$gender, function(x) 
  lapply(x[c("s100b", "ndka")], function(y) roc(y, response=x$outcome)))

# aSAH$gender: Male
# $s100b
# 
# Call:
#   roc.default(response = x$outcome, predictor = y)
# 
# Data: y in 22 controls (x$outcome Good) < 20 cases (x$outcome Poor).
# Area under the curve: 0.7727
# 
# $ndka
# 
# Call:
#   roc.default(response = x$outcome, predictor = y)
# 
# Data: y in 22 controls (x$outcome Good) < 20 cases (x$outcome Poor).
# Area under the curve: 0.5523
# 
# --------------------------------------------------- 
#   aSAH$gender: Female
# $s100b
# 
# Call:
#   roc.default(response = x$outcome, predictor = y)
# 
# Data: y in 50 controls (x$outcome Good) < 21 cases (x$outcome Poor).
# Area under the curve: 0.72
# 
# $ndka
# 
# Call:
#   roc.default(response = x$outcome, predictor = y)
# 
# Data: y in 50 controls (x$outcome Good) < 21 cases (x$outcome Poor).
# Area under the curve: 0.6671