Skip to contents

score how much activity two records show

Source: R/activity.R
score_count.Rd

Compares a per-record count – by default the `ROWCOUNT` column [transform_transaction_to_master()] produces – between the RAW and ANON sides.

Usage

score_count(
  dat_raw_anon,
  target = "ROWCOUNT",
  row_number = "ROW_NUMBER",
  method = c("log_ratio", "absolute", "relative"),
  generalized = c("stop", "warn", "ignore"),
  .fn_name = "score_count"
)

Arguments

dat_raw_anon

dataframe of raw_anon form

target

name of the count column (default `"ROWCOUNT"`)

row_number

name of the row-number column *before* the RAW_/ANON_ prefixing done by [join_raw_anon_data()] (default: "ROW_NUMBER")

method

`"log_ratio"` (default), `"absolute"` or `"relative"` (`|a - b| / max(a, b, 1)`, bounded in \[0, 1\])

generalized

what to do when `target` turns out to hold generalised values on the ANON side: `"stop"` (default), `"warn"` or `"ignore"`. A count published as a band (`"[10,20)"`) is not a number, so this only decides which of the two errors is raised.

.fn_name

name used in error messages; a function that wraps this one passes its own name so the message points at the function the user actually called

Value

a "reid_scores" table (a distance: smaller is a better match)

Details

The default `method = "log_ratio"` compares `log1p` of the two counts rather than the counts themselves, because activity counts are heavy tailed: the gap between 2 visits and 7 is strong evidence that these are different people, while the same gap between 200 and 205 is nothing. An absolute difference scores those two situations identically and so spends most of its discriminating power on the few busiest records.

Examples

tran <- create_dummy_transaction_data(people = 20, size = 4)
master <- transform_transaction_to_master(
  tran, DYNAMIC_NUM = "NUM_DYNAMIC", STATIC_NUM = "NUM_STATIC"
)
j <- join_raw_anon_data(master, master)
match_greedy(score_count(j, row_number = "ROW_NUMBER"))
#>    ANON_ROW_NUMBER RAW_ROW_NUMBER CONFIDENCE RESULT
#> 1                1              1   0.000000   TRUE
#> 2                2             24   0.000000  FALSE
#> 3                3             18   0.000000  FALSE
#> 4                4              2   0.000000  FALSE
#> 5                5              5   0.000000   TRUE
#> 6                6             10   0.000000  FALSE
#> 7                7              7   0.000000   TRUE
#> 8                8              6   0.000000  FALSE
#> 9                9              9   0.000000   TRUE
#> 10              10              6   0.000000  FALSE
#> 11              11              5   0.000000  FALSE
#> 12              13             13   0.000000   TRUE
#> 13              18             19   0.000000  FALSE
#> 14              19              5   0.000000  FALSE
#> 15              24              4   0.000000  FALSE
#> 16              28             28   1.104554   TRUE
#> 17              37              6   0.000000  FALSE
#> 18              51             51   1.160345   TRUE