sampleD

Implementation of the Stochastic Expectation Maximisation (StEM) approach to Record Linkage described in the paper by K. Robach, S. L. van der Pas, M. A. van de Wiel and M. H. Hof (2024, <doi:10.48550/arXiv.2407.06835>); see 'citation("FlexRL")' for details. This is a record linkage method, for finding the common set of records among 2 data sources based on Partially Identifying Variables (PIVs) available in both sources. It includes modelling of dynamic Partially Identifying Variables (e.g. postal code) that may evolve over time and registration errors (missing values and mistakes in the registration). Low memory footprint.

Kayané ROBACH

FlexRL

A Flexible Model for Record Linkage

sampleD function

<dl><dt>S</dt>
<dd>IntegerMatrix where each row correspond to the indices (from source A and source B) of records for which the true values matches (representing the potential links)</dd>
<dt>LLA</dt>
<dd>NumericVector gives the likelihood contribution of each non linked record from A</dd>
<dt>LLB</dt>
<dd>NumericVector gives the likelihood contribution of each non linked record from B</dd>
<dt>LLL</dt>
<dd>NumericVector gives the likelihood contribution of each potential linked records (from select)</dd>
<dt>gamma</dt>
<dd>NumericVector repeats the value of the parameter gamma (proportion of linked records) number of potential linked records (nrow of S) times</dd>
<dt>loglik</dt>
<dd>double for the value of the current complete log likelihood of the model</dd>
<dt>nlinkrec</dt>
<dd>integer for the current number of linked records</dd>
<dt>sumRowD</dt>
<dd>A LogicalVector vector indicating, for each row of the linkage matrix, i.e. for each record in the smallest file A, whether the record has a link in B or not.</dd>
<dt>sumColD</dt>
<dd>A LogicalVector vector indicating, for each column of the linkage matrix, i.e. for each record in the largest file B, whether the record has a link in A or not.</dd></dl>

Arguments

sampleD — sampleD

<dl>

<dt>S</dt>
<dd>IntegerMatrix where each row correspond to the indices (from source A and source B) of records for which the true values matches (representing the potential links)</dd>


<dt>LLA</dt>
<dd>NumericVector gives the likelihood contribution of each non linked record from A</dd>


<dt>LLB</dt>
<dd>NumericVector gives the likelihood contribution of each non linked record from B</dd>


<dt>LLL</dt>
<dd>NumericVector gives the likelihood contribution of each potential linked records (from select)</dd>


<dt>gamma</dt>
<dd>NumericVector repeats the value of the parameter gamma (proportion of linked records) number of potential linked records (nrow of S) times</dd>


<dt>loglik</dt>
<dd>double for the value of the current complete log likelihood of the model</dd>


<dt>nlinkrec</dt>
<dd>integer for the current number of linked records</dd>


<dt>sumRowD</dt>
<dd>A LogicalVector vector indicating, for each row of the linkage matrix, i.e. for each record in the smallest file A, whether the record has a link in B or not.</dd>


<dt>sumColD</dt>
<dd>A LogicalVector vector indicating, for each column of the linkage matrix, i.e. for each record in the largest file B, whether the record has a link in A or not.</dd>

</dl>

sampleD: sampleD

Description

Usage

Value

Arguments