0

I would like to apply a function to all rows of a data frame where each application the columns as distinct inputs (not like mean, rather as parameters).

I wonder what the tidy way is to do the following:

# Data
successes <- c(0,3,6,15,15,17,12,9,22,33)
trials <- c(50,1788,1876,3345,1223,856,342,214,265,257)
prognosis <- 0.01*c(0.05,0.10,0.25,0.5,0.75,1.3,2,3.4,6,10)
test_data = data.frame(successes = successes, trials = trials, 
                       prognosis = prognosis, p_value1 = NA, p_value2 = NA)

for(i in 1: nrow(test_data)){
  test_data$p_value1[i] = binom.test(test_data$successes[i], test_data$trials[i],
                                    test_data$prognosis[i], "less")$p.value
  test_data$p_value2[i] = binom.test(test_data$successes[i], test_data$trials[i],
                                     test_data$prognosis[i], "greater")$p.value
}
Richi W
  • 3,534
  • 4
  • 20
  • 39

1 Answers1

1

One possible way is this:

successes <- c(0,3,6,15,15,17,12,9,22,33)
trials <- c(50,1788,1876,3345,1223,856,342,214,265,257)
prognosis <- 0.01*c(0.05,0.10,0.25,0.5,0.75,1.3,2,3.4,6,10)
test_data = data.frame(successes = successes, trials = trials, 
                       prognosis = prognosis, p_value1 = NA, p_value2 = NA)


library(dplyr)

test_data %>%
  rowwise() %>%
  mutate(p_value1 = binom.test(successes, trials, prognosis, "less")$p.value,
         p_value2 = binom.test(successes, trials, prognosis, "greater")$p.value) %>%
  ungroup()

# # A tibble: 10 x 5
#     successes trials prognosis p_value1 p_value2
#      <dbl>  <dbl>     <dbl>    <dbl>    <dbl>
# 1        0.    50.  0.000500    0.975   1.00  
# 2        3.  1788.  0.00100     0.893   0.266 
# 3        6.  1876.  0.00250     0.806   0.330 
# 4       15.  3345.  0.00500     0.396   0.697 
# 5       15.  1223.  0.00750     0.975   0.0467
# 6       17.   856.  0.0130      0.966   0.0595
# 7       12.   342.  0.0200      0.978   0.0447
# 8        9.   214.  0.0340      0.805   0.306 
# 9       22.   265.  0.0600      0.950   0.0786
# 10       33.   257.  0.100       0.943   0.0822

Or using a vectorized function without rowwise:

# create function and vectorise it
GetPvalue = function(s, t, p, alt) {binom.test(s, t, p, alt)$p.value}
GetPvalue = Vectorize(GetPvalue)

test_data %>%
  mutate(p_value1 = GetPvalue(successes, trials, prognosis, "less"),
         p_value2 = GetPvalue(successes, trials, prognosis, "greater"))

#    successes trials prognosis  p_value1   p_value2
# 1          0     50    0.0005 0.9753038 1.00000000
# 2          3   1788    0.0010 0.8933086 0.26613930
# 3          6   1876    0.0025 0.8061877 0.32975624
# 4         15   3345    0.0050 0.3963610 0.69722243
# 5         15   1223    0.0075 0.9748903 0.04667939
# 6         17    856    0.0130 0.9656352 0.05952219
# 7         12    342    0.0200 0.9781863 0.04473155
# 8          9    214    0.0340 0.8047247 0.30581962
# 9         22    265    0.0600 0.9503332 0.07855963
# 10        33    257    0.1000 0.9433326 0.08219425
AntoniosK
  • 15,991
  • 2
  • 19
  • 32