[Question]: summarize_vars_in_cols no longer exists

[imazubi]

What is your question?

Hi Nest folks,

In older versions (< 0.8.0) of tern I was using summarize_vars_in_cols but I just saw in the later versions this function is no longer an exported object (breaking change for me). Is there any alternative so that we can summarize the statistics in columns?

This was the file
Thanks,

IZ

Code of Conduct

• I agree to follow this project's Code of Conduct.

Contribution Guidelines

• I agree to follow this project's Contribution Guidelines.

Security Policy

• I agree to follow this project's Security Policy.

[Melkiades]

We had to change its name and behavior because it was not paginating correctly. You can find the renaming commit here . It is virtually the same with the difference that it uses analyze instead of summarize which is the correct way to produce analyze rows (previously they were content rows, i.e. label rows with results). Here you can find the main change. Being it on the column space, having "label-content" does not make a lot of sense, and therefore the change from summarize_row_groups into analyze_colvars.

In practice for you, it is enough to change the name of the function into analyze_vars_in_cols and maybe some parameter name (e.g. vars instead of var). Let me know if this works for you

ps: the renaming should be in the news file, I do not know if it was correctly propagated afterward.

[imazubi]

Thanks @Melkiades .

From what I am seeing, this works so far so good.

basic_table() %>%
  split_rows_by("SEX") %>%
  analyze_vars_in_cols(
    vars = c("AGE"),
    .stats = c("n", "mean", "mean_ci", "mean_pval"),
    .labels = c("n" = "n", "mean" = "mean", "mean_ci" = "mean CI", "mean_pval" = "Mean P-value"),
    nested = FALSE
  ) %>%
  build_table(df = ex_adsl) 

However, I was trying to first get overall statistics and then split rows by two variables to get the same statistics. See the below example where I am just splitting by one variable.

basic_table() %>%
  split_rows_by("SEX") %>%
  analyze_vars_in_cols(
    vars = c("AGE"),
    .stats = c("n", "mean", "mean_ci", "mean_pval"),
    .labels = c("n" = "n", "mean" = "mean", "mean_ci" = "mean CI", "mean_pval" = "Mean P-value"),
    nested = FALSE
  ) %>%
  split_rows_by("RACE", nested = FALSE) %>%
  analyze_vars_in_cols(
    vars = c("AGE"),
    .stats = c("n", "mean", "mean_ci", "mean_pval"),
    .labels = c("n" = "n", "mean" = "mean", "mean_ci" = "mean CI", "mean_pval" = "Mean P-value"),
    nested = FALSE
  ) %>%
  build_table(df = ex_adsl) 

I tried with the nested argument but did not work. I was able to do this with the former summarize_vars_in_cols()

> packageVersion("rtables")
[1] ‘0.6.0.3’

[Melkiades]

I think what you ask for makes sense to implement again. We do not have currently any need for this mixed behavior but we should support it. I am working on it ;)

[Melkiades]

@imazubi now you can do it with something like the following. Note that the switcher between summarize and analyze is a flat called do_summarize* and you must do the split in cols only once. Also, note that every row split should have the child labels off (hidden). It is possible that we will converge on a more general solution in the future, but this allows now to do what you wanted with the plus that it can be paginated and sorted correctly.

# More nesting
  lyt <- basic_table() %>%
    split_rows_by("SEX") %>%
    analyze_vars_in_cols(
      vars = "AGE",
      do_summarize_row_groups = TRUE
    ) %>%
    split_rows_by("RACE", child_labels = "hidden", split_fun = drop_split_levels) %>%
    analyze_vars_in_cols(
      vars = "AGE",
      split_col_vars = FALSE,
      do_summarize_row_groups = TRUE
    ) %>%
    split_rows_by("ARM", child_labels = "hidden", split_fun = drop_split_levels) %>%
    analyze_vars_in_cols(
      vars = "AGE",
      split_col_vars = FALSE,
      do_summarize_row_groups = TRUE
    ) %>%
    split_rows_by("STRATA1", child_labels = "hidden", split_fun = drop_split_levels) %>%
    analyze_vars_in_cols(
      vars = "AGE",
      split_col_vars = FALSE
    )

  tbl <- testthat::expect_silent(build_table(lyt, df = tern_ex_adpp))

  # Again sorting works
  tbl_sorted <- sort_at_path(tbl, c("SEX", "*", "RACE"), cont_n_onecol(1))
  tbl_sorted <- sort_at_path(tbl_sorted, c("SEX", "*", "RACE", "*", "ARM", "*", "STRATA1"), scorefun(1))