wordfish_comparison: Wordfish Comparison.
In preText: Diagnostics to Assess the Effects of Text Preprocessing Decisions

Description Usage Arguments Value Examples

Calculated Wordfish scores for a list of dfm objects with temporal filtering.

1 2	wordfish_comparison(dfm_list, years, anchors = c(1, 24), proportion_threshold = 1, document_inidices = NULL)

`dfm_list`	A list of quanteda dfm objects generated by the 'factorial_preprocessing()' and returned in the '$dfm_list' field
`years`	A numeric vector giving the year for each document.
`anchors`	A numeric vector of length two used to anchor the Wordfish estimates. Defaults to c(1,24) which should work for the UK parliament docs.
`proportion_threshold`	proportion of years a term must be included in to be included in the Wordfish analysis.
`document_inidices`	An option vector of row indices to be used. Useful for using a subset of the data for analysis.

A result list object

## Not run: 
# replicates wordfish aanalysis from Denny and Spirling (2016)
# load the package
library(preText)
# load in the data
data("UK_Manifestos")
# preprocess data
preprocessed_documents <- factorial_preprocessing(
    UK_Manifestos,
    use_ngrams = TRUE,
    infrequent_term_threshold = 0.02,
    verbose = TRUE)
# get the years each document was written and store them as a numeric vector
dfm <- preprocessed_documents$dfm_list[[1]]
rl <- function(str) {
    stringr::str_replace_all(str,"[A-Za-z]+","")
}
years <- as.numeric(sapply(rownames(dfm),rl))

# use the wordfish_comparison function to compare all dfms. We are using
# conservative and labour manifestos from 1983, 1987, 1992, and 1997 for a total
# of 8 manifestos. These are indicated by the document_inidices = c(19:22,42:45)
# argument. You can see the document names by entering rownames(dfm) into the
# console. We need to set the anchors to 5,1 because anchoring is applied in the
# reduced dfm. We are also only including terms that appear atleast once in a
# manifesto from each of the 4 years, to deal with the strong temporal effects.
wordfish_results <- wordfish_comparison(
    preprocessed_documents$dfm_list,
    years,
    anchors = c(1,5),
    proportion_threshold = 1,
    document_inidices = c(19:22,42:45))

## End(Not run)