Suggested a fix in defining study code in the createStudyTable function

[g-antonello]

In the previous implementation the study code in createStudyTable did not account for the following edge cases:

• Same first author name AND same year, but different publication
• Special characters that the regex could not capture (case of PMID 34884399)

This resulted in a different number of lines in the study tables (run code example attached to verify).

Additionally, I have added a parameter to the function to allow the user to arbitrarily add other columns as they think best, with the includeAlso parameter.

bsdb.df <- bugsigdbr::importBugSigDB(cache = FALSE)

# Some studies have NAs on PMID, URL and DOI, but not in all 3
table(rowSums(is.na(select(bsdb, PMID, DOI, URL))), useNA = "always")

# unique studies based on their various links  
unique_links <- unique(paste(bsdb$PMID, bsdb$DOI, bsdb$URL))
length(unique_links)


# old function as in v 0.99.5

createStudyTable_old <-function(dat){
  studies <- data.frame(Study=paste0(str_extract(dat$Authors, "[A-Za-z]+[:space:]"), dat$Year),
                        Condition=dat$Condition,
                        Cases=dat$`Group 1 sample size`,
                        Controls=dat$`Group 0 sample size`,
                        `Study Design`=dat$`Study design`)
  studies %>% group_by(Study) %>% summarize(Condition=first(Condition), 
                                            Cases=max(Cases),
                                            Controls=max(Controls), 
                                            `Study Design`=first(`Study.Design`))
  
}

# newly proposed function

createStudyTable_new <- function(bsdb.df, includeAlso = NULL) {
  # input check
  if (!is_null(includeAlso)) {
    if (!all(includeAlso %in% colnames(bsdb.df))) {
      stop(paste(
        "The following columns are not found in the input data frame:",
        paste(includeAlso[!(includeAlso %in% colnames(bsdb.df))], collapse = ", ")
      ))
    }
  }
  # Core of the change is in how study IDs are generated, see function in 
  # simple.R. NB: the function also fixes DOI links as side effect, now. 
  bsdb_with_StudyCodes.df <- .make_unique_study_ID(bsdb.df)
  
  # some dplyr-fu to summarize tables, with more recent syntax
  study_table_fixed <- bsdb_with_StudyCodes.df %>%
    group_by(`Study code`) %>%
    reframe(
      MaxCases = max(`Group 1 sample size`),
      MaxControls = max(`Group 0 sample size`),
      across(
        all_of(
          c("Study design", "Condition", "PMID", "DOI", "URL", includeAlso)
        ),
        .fns = function(x)
          paste(unique(x), collapse = "; ")
      ),
      N_signatures = n()
    ) %>%
    relocate(N_signatures, .after = Condition)
  
  return(study_table_fixed)
}

study_table_old <- createStudyTable_old(bsdb.df)

study_table_new <- createStudyTable_new(bsdb.df)

dim(study_table_old)

dim(study_table_new)

[cmirzayi]

I think this is a really cool improvement. Thank you. I figure we can merge it in.