How do you ignore the "NS" and just show those which are significantly different? about ggsignif

[const-ae]

There is currently no way to plot only brackets for significant differences. The idea is that you decide in the beginning which comparisons are interesting for you and this package helps you to show their comparison.

[aleysia]

There's a way to avoid printing "NS" for non-significant comparisons, but this won't prevent the lines from being drawn. Since you can define what symbols to use at each significance level, something like

map_signif_level=c("***"=0.001,"**"=0.01, "*"=0.05, " "=2)

will cause anything with p>0.05 to not print any text.

[Midnighter]

What I did is to filter the data I pass to the annotations argument, for example, a label column, by p-value and set the uninteresting ones to NA. Then they will not be drawn at all. (Maybe this was not possible in earlier versions.)

[asierFernandezP]

Hi @Midnighter, could you please share how you did that? Thank you!

[Midnighter]

Similarly to the advanced example at the bottom of https://const-ae.github.io/ggsignif/

I created a data frame using the pairwiseComparisons::pairwise_comparisons function similarly to the following where group is the variable for the comparison groups and measure are the values.

annotation_df <- tbl %>%
      pairwiseComparisons::pairwise_comparisons(
        group,
        measure
      ) %>%
      dplyr::mutate(
        groups = purrr::pmap(
          .l = list(group1, group2), .f = c
        ),
        label = round(p.value, 3)
      ) %>%
      dplyr::arrange(group1) %>%
      dplyr::filter(p.value < 0.05)

I then create a ggplot to my liking and directly use the above table.

plot <- ggplot2::ggplot(...) +
    ...

if (nrow(annotation_df) > 0) {
    plot + ggsignif::geom_signif(
          comparisons = annotation_df$groups,
          annotations = annotation_df$label,
          na.rm = TRUE,
          parse = TRUE
        )
}

I've put this into convenient functions that I use many times in my project. That makes it a bit less painful to prepare the data in the right way.