diyar (version 0.5.1)

link_wf: Record linkage

Description

Deterministic and probabilistic record linkage Assign unique identifiers to records based on partial, nested or calculated probabilities.

Usage

links_af_probabilistic(
  attribute,
  blocking_attribute = NULL,
  cmp_func = diyar::exact_match,
  attr_threshold = 1,
  probabilistic = TRUE,
  m_probability = 0.95,
  u_probability = NULL,
  score_threshold = 1,
  repeats_allowed = FALSE,
  permutations_allowed = FALSE,
  data_source = NULL,
  ignore_same_source = TRUE,
  display = "none"
)
links_wf_probabilistic(
  attribute,
  blocking_attribute = NULL,
  cmp_func = diyar::exact_match,
  attr_threshold = 1,
  probabilistic = TRUE,
  m_probability = 0.95,
  u_probability = NULL,
  score_threshold = 1,
  id_1 = NULL,
  id_2 = NULL,
  return_weights = FALSE,
  ...
)
prob_score_range(attribute, m_probability = 0.95, u_probability = NULL)

Value

pid; list

Arguments

attribute

[atomic|list|data.frame|matrix|d_attribute]. Attributes to compare.

blocking_attribute

[atomic]. Passed to criteria in links.

cmp_func

[list|function]. String comparators for each attribute. See Details.

attr_threshold

[list|numeric|number_line]. Weight-thresholds for each cmp_func. See Details.

probabilistic

[logical]. If TRUE, scores are assigned base on Fellegi-Sunter model for probabilistic record linkage. See Details.

m_probability

[list|numeric]. The probability that a matching records are the same entity.

u_probability

[list|numeric]. The probability that a matching records are not the same entity.

score_threshold

[numeric|number_line]. Score-threshold for linked records. See Details.

repeats_allowed

[logical] Passed to repeats_allowed in links.

permutations_allowed

[logical] Passed to permutations_allowed in links.

data_source

[character]. Passed to data_source in links.

ignore_same_source

[logical] Passed to ignore_same_source in links.

display

[character]. Passed to display in links.

id_1

[list|numeric]. Record id or index of one half of a record-pair.

id_2

[list|numeric]. Record id or index of one half of a record-pair.

return_weights

If TRUE, returns the match-weights and score-thresholds for record pairs.

...

Arguments passed to links

Details

links_wf_probabilistic() - A wrapper function of links with a with a specific sub_criteria and to achieve to achieve probabilistic record linkage It excludes functionalities for the nested and multi-stage linkage. links_wf_probabilistic() requires a score_threshold in advance. To help with this, prob_score_range() can be used to return the range of scores attainable for a given set of attribute, m and u-probabilities. Additionally, id_1 and id_2 can be used to link specific records pairs, aiding the review of potential scores.

links_af_probabilistic() - A simpler version of links. It excludes functionalities for the batched, nested and multi-stage linkage. links_af_probabilistic() requires a score_threshold in advance, however, since it exports the match weights, the score_threshold can be changed after the analysis.

References

Fellegi, I. P., & Sunter, A. B. (1969). A Theory for Record Linkage. Journal of the Statistical Association, 64(328), 1183 - 1210. https://doi.org/10.1080/01621459.1969.10501049

Asher, J., Resnick, D., Brite, J., Brackbill, R., & Cone, J. (2020). An Introduction to Probabilistic Record Linkage with a Focus on Linkage Processing for WTC Registries. International journal of environmental research and public health, 17(18), 6937. https://doi.org/10.3390/ijerph17186937. See vignette("links") for more information.

See Also

links

Examples

data(patient_records)
# Weighted (probabilistic) comparison of forename, middlename and surname
criteria_1 <- as.list(patient_records[c("forename", "middlename", "surname")])

# Possible scores when m-probability is 0.95
prob_scores <- prob_score_range(attribute = criteria_1,
                                m_probability = 0.95,
                                u_probability = NULL)
if (FALSE) {
# Probabilistic record linkage with 'links_af_probabilistic()'
pids_1a <- links_af_probabilistic(attribute = criteria_1,
                        cmp_func = exact_match,
                        attr_threshold = 1,
                        probabilistic = TRUE,
                        m_probability = 0.95,
                        score_threshold = prob_scores$mid_scorce,
                        display = "stats")

# Equivalent with 'links_wf_probabilistic()'
pids_1b <- links_wf_probabilistic(attribute = criteria_1,
                                  cmp_func = exact_match,
                                  attr_threshold = 1,
                                  probabilistic = TRUE,
                                  m_probability = 0.95,
                                  score_threshold = prob_scores$mid_scorce,
                                  display = "progress",
                                  recursive = TRUE,
                                  check_duplicates = TRUE)

# Less thorough but faster equivalent with `links_wf_probabilistic()`
pids_1c <- links_wf_probabilistic(attribute = criteria_1,
                                  cmp_func = exact_match,
                                  attr_threshold = 1,
                                  probabilistic = TRUE,
                                  m_probability = 0.95,
                                  score_threshold = prob_scores$mid_scorce,
                                  display = "progress",
                                  recursive = FALSE,
                                  check_duplicates = FALSE)

# Each implementation can lead to different results
summary(pids_1a$pid)
summary(pids_1b$pid)
summary(pids_1c$pid)
}

# Weighted (non-probabilistic) comparison of forename, middlename and age difference
criteria_2 <- as.list(patient_records[c("forename", "middlename", "dateofbirth")])
age_diff <- function(x, y){
  diff <- abs(as.numeric(x) - as.numeric(y))
  wgt <-  diff %in% 0:(365 * 10) & !is.na(diff)
  wgt
}

pids_2a <- links_af_probabilistic(attribute = criteria_2,
                        blocking_attribute = patient_records$surname,
                        cmp_func = c(exact_match, exact_match, age_diff),
                        score_threshold = number_line(3, 5),
                        probabilistic = FALSE,
                        display = "stats")

# Larger weights can be assigned to particular attributes through `cmp_func`
# For example, a smaller age difference can contribute a higher score (e.g 0 to 3)
age_diff_2 <- function(x, y){
  diff <- as.numeric(abs(x - y))
  wgt <-  diff %in% 0:(365 * 10) & !is.na(diff)
  wgt[wgt] <- match(as.numeric(cut(diff[wgt], 3)), 3:1)
  wgt
}
pids_2b <- links_af_probabilistic(attribute = criteria_2,
                        blocking_attribute = patient_records$surname,
                        cmp_func = c(exact_match, exact_match, age_diff_2),
                        score_threshold = number_line(3, 5),
                        probabilistic = FALSE,
                        display = "stats")

head(pids_2a$pid_weights, 10)
head(pids_2b$pid_weights, 10)