Description Usage Arguments Details Value Examples

View source: R/code_for_paper.r

Function to impute missing agreement patterns and then to link data

1 2 |

`d` |
Matrix of agreement patterns with final column counting the number of times that pattern was observed. See Details |

`initial_m` |
starting probabilities for per-field agreement in record pairs, both records being generated from the same individual. Defaults to NULL |

`initial_u` |
starting probabilities for per-field agreement in record pairs, with the two records being generated from differing individuals Defaults to NULL |

`p_init` |
starting probability that both records for a randomly selected record pair is associated with the same individual |

`fixed_col` |
vector indicating columns that are not to be updated in initial EM algorithm. Useful if good prior estimates of the mis-match probabilities. See details |

`alg` |
character; see Details |

`d`

is a numeric matrix with N rows corresponding to N record pairs, and L+1 columns the first L of which show the field agreement patterns observed over the record pairs, and the last column the total number of times that pattern was observed in the database. The code 0 is used for a field that differs for two record, 1 for a field that agrees, and 2 for a missing field. `fixed_col`

indicates the components of the `u`

vector (per field probabilities of agreement for 2 records from differing individuals) that are not to be updated when applying the EM algorithm to estimate components of the Feligi Sunter model. `alg`

has four possible values. The default `'m'`

fits a log-linear model for the agreement counts only within the record pairs that corresponds to the same individual, `'b'`

fits differing log-linear models for the 2 clusters, `'i'`

corresponds to the original Feligi Sunter algorithm, with probabilities estimated via the EM algorithm, `'a'`

fits all the previously listed models

A list, the first component is a matrix - the posterior probabilities of being a true match is the last column, the second component are the fitted models used to generate the predicted probabilities

1 2 3 4 5 6 7 8 9 10 11 | ```
# Simulate data
m_probs <- rep(0.8,6)
u_probs <- rep(0.2,6)
means_match <- -1*qnorm(1-m_probs)
means_mismatch <- -1*qnorm(1-u_probs)
missingprobs <- rep(.2,6)
thedata <- do_sim(cor_match=0.2,cor_mismatch=0,nsample=10^4,pi_match=.5,
m_probs=rep(0.8,5),u_probs=rep(0.2,5),missingprobs=rep(0.4,5))
colnames(thedata) <- c(paste("V",1:5,sep="_"),"count")
output <- linkd(thedata)
output$fitted_probs
``` |

Embedding an R snippet on your website

Add the following code to your website.

For more information on customizing the embed code, read Embedding Snippets.