R/analysis_LDA.R

In weecology/MATSS-LDATS: LDA_TS meta-analysis using MATSS

#' @name run_LDA
#' @title Run Latent Dirichlet Allocation on Tabular Data
#' @description Test the Latent Dirichlet Allocation (LDA) model on the data
#'   with different number of topics (from 2 to `max_topics`), select the best
#'   one using AIC, and return the model object that is selected.
#'
#' @param data a data.frame or tibble; each row is an observation (e.g. in
#'   time or space), and each column is a variable. Here, the common usage is
#'   for each column to be a species or taxon, and each row to be an observed
#'   sample. In the original specification for LDA, each row is a document,
#'   and each column is a word, with the entries being the counts of the words
#'   in each document.
#' @param max_topics the maximum number of topics to try (the function will
#'   test a number of topics from 2 to `max_topics`)
#' @inheritParams LDATS::LDA_set
#'
#' @return the best fit model object, from running `LDATS::parLDA()`
#' @export
#'
run_LDA <- function(data,
                    max_topics = 6, nseeds = 200,
                    control = list())
{
    if (!MATSS::check_data_format(data))
    {
        wrongFormat = simpleWarning("Incorrect data structure, see data-formats vignette")
        tryCatch(warning(wrongFormat), finally = return('Incorrect data structure'))
    }

    #### Run LDAs ####
    abundances <- data$abundance
    topics_vector <- seq(from = 2, to = max_topics, by = 1)
    LDA_models = LDATS::LDA_set(document_term_table = abundances,
                                topics = topics_vector,
                                nseeds = nseeds, control = control)

}