[Feature Request]: Add weights option to ancova

[vpilliano]

Feature description

In summarize_ancova.R

Can you please add a "weights" option to be passed to emmeans, just like you did for MMRM?

This parameter needs to be added to h_ancova, and every function calling it (s_ancova, a_ancova, and summarize_ancova)

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.

[yurovska]

Wondering, for an extremely urgent request where I need weights = "proportional", would it be enough to monkey-patch h_ancova as follows:

# Modified tern::h_ancova to make weights to be "proportional"

h_ancova_proportional <- function(.var,
                                  .df_row,
                                  variables,
                                  interaction_item = NULL) {

  <Copy/paste the code of h_ancova>

  emmeans_fit <- emmeans::emmeans(
    lm_fit,
    specs = specs,
    data = .df_row,
    weights = "proportional"   # Added just one line
  )
  
  emmeans_fit
}

# Replace h_ancova with the modified one in tern's namespace
# prior to use of summarize_ancova

environment(h_ancova_proportional) <- asNamespace("tern")
assignInNamespace("h_ancova", h_ancova_proportional, ns = "tern")

[vpilliano]

Wondering, for an extremely urgent request where I need weights = "proportional", would it be enough to monkey-patch h_ancova as follows:

Modified tern::h_ancova to make weights to be "proportional"

h_ancova_proportional <- function(.var,
.df_row,
variables,
interaction_item = NULL) {

<Copy/paste the code of h_ancova>

emmeans_fit <- emmeans::emmeans(
lm_fit,
specs = specs,
data = .df_row,
weights = "proportional" # Added just one line
)

emmeans_fit
}

Replace h_ancova with the modified one in tern's namespace

prior to use of summarize_ancova

environment(h_ancova_proportional) <- asNamespace("tern")
assignInNamespace("h_ancova", h_ancova_proportional, ns = "tern")

Hi Andrii,

That could work.
My workaround was to replace all the functions up to summarize_ancova

custom_ancova <- function(df,
                          .var,
                          .df_row,
                          variables,
                          .ref_group,
                          .in_ref_col,
                          conf_level) {
 
  
  emmeans_fit <- h_ancova_proportional(.var = .var, variables = variables, .df_row = .df_row)
...

a_custom_ancova <- make_afun(
  custom_ancova,
  .indent_mods = c("n" = 0L, "lsmean" = 0L, "lsmean_diff" = 0L, "lsmean_diff_ci" = 1L, "pval" = 1L),
  .formats = c(
    "n" = "xx",
    "lsmean" = "xx.xx",
    "lsmean_ci" = "(xx.xx, xx.xx)",
    "lsmean_diff" = "xx.xx",
    "lsmean_diff_ci" = "(xx.xx, xx.xx)",
    "pval" = "x.xxxx | (<0.0001)"
  ),
  .null_ref_cells = FALSE
)
new_summ_ancova <- function (lyt, vars, var_labels, ..., show_labels = "visible", 
                             table_names = vars, .stats = NULL, .formats = NULL, .labels = NULL, 
                             .indent_mods = NULL) 
{
  afun <- make_afun(a_custom_ancova, .stats = .stats, .formats = .formats, 
                    .labels = .labels, .indent_mods = .indent_mods)
  analyze(lyt, vars, var_labels = var_labels, show_labels = show_labels, 
          table_names = table_names, afun = afun, extra_args = list(...))
}

[yurovska]

@vpilliano Oh, you probably also modified s_ancova in order to include lsmean_ci, which is not available there for no reason. Could you please share the full code of your custom_ancova?

[vpilliano]

Yes, I modified that too.
It's quite messy and full of ad-hoc edits for my use case so it's not going to help you, unfortunately

[Melkiades]

Wondering, for an extremely urgent request where I need weights = "proportional", would it be enough to monkey-patch h_ancova as follows:

Modified tern::h_ancova to make weights to be "proportional"

h_ancova_proportional <- function(.var,
.df_row,
variables,
interaction_item = NULL) {

<Copy/paste the code of h_ancova>

emmeans_fit <- emmeans::emmeans(
lm_fit,
specs = specs,
data = .df_row,
weights = "proportional" # Added just one line
)

emmeans_fit
}

Replace h_ancova with the modified one in tern's namespace

prior to use of summarize_ancova

environment(h_ancova_proportional) <- asNamespace("tern")
assignInNamespace("h_ancova", h_ancova_proportional, ns = "tern")

I think this would be a simple PR. Could you try adding it? it is enough to add the weights variable up to the main function