In `total_scores` add option to drop variables that were aggregated

[mark-andrews]

I think may be a relatively simple manner of deselecting by the selectors.

[Mark-Torrance]

This?

total_scores <- function(.data, ..., .method = 'mean', .append = FALSE, .drop = FALSE){

  totalling_function <- function(.data_selection, .method){
    switch(.method,
           mean = rowMeans(.data_selection, na.rm = T),
           sum = rowSums(.data_selection, na.rm = T),
           sum_like = rowMeans(.data_selection, na.rm = T) * ncol(.data_selection)
    )
  }

  if (!(.method %in% c('mean', 'sum', 'sum_like'))) {
    stop('The function that calculates the total must be either "mean", "sum" or "sum_like".')
  }

  selection_sets <- rlang::enquos(...)

  results_df <- purrr::map_dfc(selection_sets, 
                               function(selection_set){
                                 totalling_function(select(.data, !!selection_set),
                                                    .method)
                               }
  )

  if (.append) {
    Df <- .data
    if (.drop) {
      for(set in selection_sets){
        Df <- select(Df, !all_of(!!set))
      }
    } 

   bind_cols(Df, results_df) 

    } else {
    results_df
    } 
}

This...

Df_all <- read_csv("http://data.ntupsychology.net/psychometrics_demo_data.csv")

head(Df_all %>% total_scores(anxiety = starts_with('anxiety_'),
                             depression = starts_with('depression_'),
                             efficacy = starts_with('efficacy_'),
                             sociability = starts_with('sociability_'),
                             stress = starts_with('stress'),
                             .append = T, .drop = T))

gives...

gender	age	anxiety	depression	efficacy	sociability	stress
1	19	1.8	1.8	2.8	2.2	1.0
1	22	2.4	2.8	2.3	3.3	2.3
2	20	1.7	2.6	3.0	3.5	1.9
2	20	2.2	2.9	3.6	2.3	1.6
2	19	1.7	2.6	3.3	2.8	2.0
2	21	2.0	2.1	2.7	2.6	2.0

[mark-andrews]

• I implemented a variant of this.
• Having consulted Holy Writ, it is written Thou shalt not use for loops any more in R, so I used purrr::reduce to replace the for loop where the selected sets of variables are dropped.
• Added tests, as one should
• Updated documentation
• Pushed to development branch. Keep main stable and inline with version on CRAN. Everything on development will be merged or rebased into main eventually, but no great rush. Merging will happen at next CRAN update.
• See help pages with live example (last example) here: https://mark-andrews.github.io/psyntur/reference/total_scores.html
• See code here: https://github.com/mark-andrews/psyntur/blob/2e8da4849d3ac2b1d4b91a5f6aa1b1009360f031/R/psychometrics.R#L76-L113
• See commit here: https://github.com/mark-andrews/psyntur/commit/2e8da4849d3ac2b1d4b91a5f6aa1b1009360f031

[mark-andrews]

As a final point, before I close this, I think that the default should be .append = TRUE, but I am not sure if the default should be .drop = TRUE. I am inclining to .drop = TRUE by default because the total_scores operation is intended for a part of a typical psychometrics-y data analysis workflow whereby sets of items are aggregated over and then only the results of the aggregation are, typically, then used in the remaining workflow. That would entail both .append = TRUE and .drop = TRUE.