Feature request: Two or more levels of column spanners

[martinschlund]

---

Hello, and thank you for at very good package. One feature I miss is the ability to have 2 levels of column spanners. When reporting results from competing risks analysis you often want to report uni- and multivariable hazard ratios for both the event of interest and the competing risk side by side in the same table. Like this generated with excel:

example.pdf

When trying to achieve the same in R with gt() and table_spanner() the second argument overwrites the first however:

example <- tibble(predictors = c('Age', 'Gender', 'COPD'), eventofinterest_uni_hr = c('1.01 (1.00 – 1.02)', '0.81 (0.65 - 1.02)', '1.77 (1.40 - 2.24)'),
                  eventofinterest_uni_p = c(0.19, 0.07, 0.003), eventofinterest_multi_hr = c('1.01 (1.00 - 1.02)', '0.83 (0.66 - 1.06)', '1.55 (1.20 - 2.00)'),
                  eventofinterest_multi_p = c(0.2, 0.14, 0.0004), competing_uni_hr = c('1.04 (1.02 – 1.06)', '1.02 (0.68 – 1.54)', '1.39 (0.91 – 2.13)'), 
                  competing_uni_p = c(0.007, 0.91,0.12), competing_multi_hr = c('1.04 (1.02 – 1.06)', '1.06 (0.70 – 1.61)', '1.41 (0.90 – 2.21)'), 
                  competing_multi_p = c(0.0025, 0.77, 0.14))

example %>% 
  gt() %>% 
  cols_label(eventofinterest_uni_hr = 'Hazard Ratio', 
             eventofinterest_multi_hr = 'Hazard Ratio', 
             competing_uni_hr = 'Hazard Ratio', 
             competing_multi_hr = 'Hazard Ratio', 
             eventofinterest_uni_p = 'p-value', 
             eventofinterest_multi_p = 'p-value', 
             competing_uni_p=  'p-value', 
             competing_multi_p = 'p-value') %>% 
  tab_spanner(label= 'Event of Interest', 
              columns = vars(eventofinterest_uni_hr, eventofinterest_uni_p, eventofinterest_multi_hr, eventofinterest_multi_p)) %>%
  tab_spanner(label= 'Competing Risk of Death', columns = vars(competing_uni_hr, competing_uni_p, competing_multi_hr, competing_multi_p)) %>% 
  tab_spanner(label = 'Univariable', columns = vars(eventofinterest_uni_hr, eventofinterest_uni_p, competing_uni_hr, competing_uni_p), gather = F) %>% 
  tab_spanner(label = 'Multivariable', columns = vars(eventofinterest_multi_hr, eventofinterest_multi_p, competing_multi_hr, competing_multi_p), gather = F)

Would that be possible to implement in the future? I know I can use group_by() or the groupname_col- argument in gt() to get to tables on top of each other for the event of interest and the competing risk respectively. Side-by-side is however how most journals want these reported and I also think it gives a better overview of data.

Best regards

Martin Schønemann-Lund

[rich-iannone]

I agree that currently having only a single level of column spanners is a limitation. The intention is to enable multiple levels (and this will be a possibility for row groups as well). After a few more API upgrades and quality-of-life improvements, we should be in a better place to begin this work.