I' am trying to understand the logic of ideological scaling. I have a dataset consisting of monetary and fiscal policy related texts with a dimension t (=time), and j (institution). I would like to scale the texts using wordfish.
example of quanteda package:
irish_dfm <- dfm(data_corpus_irishbudget2010, remove_punct = TRUE)
wf <- textmodel_wordfish(irish_dfm, dir = c(6,5))
summary(wf)
The dir = c(6,5) setting is not clear to me.
At https://www.rdocumentation.org/packages/quanteda/versions/1.3.4/topics/textmodel_wordfish
it is stated that
"set global identification by specifying the indexes for a pair of documents such that ^θdir[1]<^θdir[2]."
Does it mean I have to pick two documents and read and impose that theta of doc 1 is lower that theta of doc 2?
