Challenge 10

purrr

For this challenge, I’ll be using the function I wrote in challenge 9 that calculates z scores and apply it multiple times.

A common use of z-scores is in anomaly detection. In this practice, we compare the most recent value in a sequence to all the values that came before to see if that value is an anomaly or not.

We can use a built in dataset for this, called airquality, which stores time series air quality information:

data(airquality)
head(airquality)

ABCDEFGHIJ0123456789

	Ozone <int>	Solar.R <int>	Wind <dbl>	Temp <int>	Month <int>	Day <int>
1	41	190	7.4	67	5	1
2	36	118	8.0	72	5	2
3	12	149	12.6	74	5	3
4	18	313	11.5	62	5	4
5	NA	NA	14.3	56	5	5
6	28	NA	14.9	66	5	6

6 rows

Our original function looks as follows:

z_score <- function(baseline, value){
  mean <- mean(baseline)
  sd <- sd(baseline)
  z_score <- abs((value - mean)/sd)
  
  results = tibble(mean = mean, sd = sd, input_value = value, z_score = z_score)
  return(results)
}

We can rewrite this so that it determines the baseline and value itself, and instead takes a vector as input:

z_score <- function(vec){
  baseline = vec %>% head(-1)
  value = vec %>% tail(1)
  
  mean <- mean(baseline, na.rm=TRUE)
  sd <- sd(baseline, na.rm=TRUE)
  z_score <- abs((value - mean)/sd)
  
  results = tibble(baseline_mean = mean, baseline_sd = sd, most_recent_value = value, z_score = z_score)
  return(results)
}

Running this on a single column we get:

z_score(airquality$Temp)

ABCDEFGHIJ0123456789

baseline_mean <dbl>	baseline_sd <dbl>	most_recent_value <int>	z_score <dbl>
77.94737	9.462221	68	1.051272

1 row

We can use purrr::map to compute this for multiple columns and join them into a single dataframe:

cols = list(airquality$Ozone, airquality$Wind, airquality$Temp)

map_dfr(cols, z_score)

ABCDEFGHIJ0123456789

baseline_mean <dbl>	baseline_sd <dbl>	most_recent_value <dbl>	z_score <dbl>
42.321739	33.066798	20.0	0.6750499
9.947368	3.532403	11.5	0.4395397
77.947368	9.462221	68.0	1.0512720

3 rows