CODEWITHSUNDEEP
rstatisticstidyverse
Timothy Mcwilliams
Timothy Mcwilliams

Reputation: 316

generate row combinations then apply a function to each row's values

I have a dataframe and my goal here is to take all possible combinations of the rank column and for each rank combination apply a function that will use the volume_metric and kpi_metric values. The resulting output would be a matrix just like the cor() function provides except with each rank combinations p-values.

Basically I want to take the first row values of volume_metric & kpi_metric and then the second row values of volume_metric & kpi_metric and apply the zTest function to them. Then 1->3, 1->4, etc.

rank <- c('ad 1', 'ad 2', 'ad 3', 'ad 4', 'ad 5', 'ad 6', 'ad 7', 'ad 8')
volume_metric <- c(12321, 12321, 1232121, 4343, 14333, 52323, 234532, 2322)
kpi_metric <- c(12, 32, 111, 334, 653, 343, 232, 212)
# The df
df <- tibble(rank, volume_metric, kpi_metric)

# A tibble: 8 x 3
   rank volume_metric kpi_metric
     <chr>         <dbl>      <dbl>
1     ad 1         12321         12
2     ad 2         12321         32
3     ad 3       1232121        111
4     ad 4          4343        334
5     ad 5         14333        653
6     ad 6         52323        343
7     ad 7        234532        232
8     ad 8          2322        212

# z-test fucntion
zTest <- function(volume1, volume2, kpi1, kpi2) {
  z_test <- prop.test(
    x=c(kpi1, kpi2),
    n=c(volume1, volume2),
    alternative = "greater",
    conf.level = 0.95,
    correct = FALSE
  )
  p_value <- z_test$p.value
  return(p_value)
}

So far I have been able to get all of the rank combinations using

possible_combinations <- combn(nrow(df), 2)

which will provide a matrix with all of the combos (the rank will always be the same as now(df)).

I tried to loop through that matrix and then subset the df but that resulted in a never ending loop 🤦🏻‍♂️.

My question here is, how do I use that matrix with the combos to index against my df and apply the zTest function, or am I thinking about this all wrong?

Upvotes: 0

Views: 38

Answers (1)

Ronak Shah
Ronak Shah

Reputation: 388982

combn accepts a function so you may pass the row number values to them and subset specific volume_metric and kpi_metric from df and pass it to zTest function.

zTest <- function(volume, kpi) {
  z_test <- prop.test(
    x=kpi,
    n=volume,
    alternative = "greater",
    conf.level = 0.95,
    correct = FALSE
  )
  p_value <- z_test$p.value
  return(p_value)
}

do.call(rbind, combn(nrow(df), 2, function(x) 
       data.frame(row1 = x[1], row2 = x[2], 
                  cor = zTest(df$volume_metric[x], df$kpi_metric[x])), 
       simplify = FALSE))

#   row1 row2        cor
#1     1    2  9.987e-01
#2     1    3  4.628e-23
#3     1    4  1.000e+00
#4     1    5  1.000e+00
#5     1    6  1.000e+00
#6     1    7  5.209e-01
#7     1    8  1.000e+00
#...

Upvotes: 2

Related Questions