Simple dyplyr grouping and counting

It is simple to get counts by groups with a familiar friend, the {dplyr) package.
The main functions you will use are

• tally() counts all rows, like n()
• count(catvars) groups by categorical variable arguments, then counts rows within groups, then ungroups. Essentially, count() combines group_by() and tally() and ungroup() into one function.
• add_tally() combines mutate() and tally() to add an new column, named n, to your dataframe.
• add_count combines mutate() and count() to add a new column to your dataframe.

Examine the code chunk below to see an example of counts from the blood dataset by African-American status, Gleason Score category, and Tumor Volume category. You can page through the results to see how many n are in each category. Adding a tally gives you the total n for your denominator.
You can edit the code to change the name of the tally column, with the argument name = "total". Give this a try. You can edit the code to sort the categories in descending order of n, with the argument sort = TRUE (after a comma) in the count function. Give this a try.

blood %>% 
  count(aa, gleason, tumor_vol) %>% 
  add_tally()

## Storing counts in `nn`, as `n` already present in input
## ℹ Use `name = "new_name"` to pick a new name.

Improvements

You can change the name of the tally column, with the argument name = "total". You can also sort the categories in descending order of n, with the argument sort = TRUE (after a comma) in the count function. See the example below

blood %>% 
  count(aa, gleason, tumor_vol, sort = TRUE) %>% 
  add_tally(name = "Total")

licorice 

licorice %>% 
  count(gender)

licorice %>% 
  count(gender, preOp_asa, smoking)

licorice %>% 
  count(gender, preOp_asa, smoking, sort = TRUE) %>% 
  add_tally(name = "big_N")

supra 

supra %>% 
  count(gender)

supra %>% 
  count(gender, group, bmi_cat6)

supra %>% 
  count(gender, group, bmi_cat6, sort = TRUE) %>% 
  add_tally(name = "Denominator")

supra %>% 
  count(gender, group, bmi_cat6, sort = TRUE) %>% 
  add_tally(name = "Denominator") %>% 
  gt() %>% 
   tab_header(
    title = "Counts by Group",
    subtitle = "From the Supraclavicular Dataset"
  ) %>% 
  tab_options(
   table.background.color = "darkorchid",
   heading.background.color = "yellow"
  ) 

supra %>% 
  count(gender, group, bmi_cat6, sort = TRUE) %>% 
  add_tally(name = "Denominator") %>% 
  flextable() %>% 
  autofit() %>% 
  bold(part = "header") %>% 
  fontsize(part = "header", size = 16) %>% 
  fontsize(part = "body", size = 14) %>% 
  align(align = "center", part = "all") %>% 
  #note conditional formatting in rows when n>7, turn column n purple and bold. Could do this for P <0.05, etc.
  color(color = "purple", i =~ n>7, j =~ n) %>% 
  bold(i =~ n>7, j =~ n)

Simple janitor tabyl with percentages

We can make this fancier with percentages, rather than just counts. See the example below to see how this is done.

blood %>% 
  tabyl(aa, tumor_vol) %>% 
  adorn_percentages() %>% 
  adorn_pct_formatting() %>% 
  adorn_ns()

licorice

licorice %>% 
  tabyl(gender, smoking)

licorice %>% 
  tabyl(gender, smoking) %>% 
  adorn_percentages() %>% 
  adorn_pct_formatting() %>% 
  adorn_ns()

{janitor} tabyl with Totals and Titles

Here is an example of how to improve a simple 2 x 2 table with the {janitor} package function tabyl by adding totals to rows/columns and titles.

supra %>% 
  tabyl(gender, group) %>% 
  adorn_totals(where = c("row", "col")) %>% 
  adorn_title(row_name = "Gender", 
              col_name = "Group Assigned",
              placement = "combined")

Note that the adorn_title() placement argument can also be “top” - see what this looks like in the chunk below.

supra %>% 
  tabyl(gender, group) %>% 
  adorn_totals(where = c("row", "col")) %>% 
  adorn_title(row_name = "Gender", 
              col_name = "Group Assigned",
              placement = "top")

blood

blood %>% 
  tabyl(fam_hx, recur)

blood %>% 
  tabyl(fam_hx, recur) %>% 
  adorn_totals(where = c("row", "col")) %>% 
  adorn_title(row_name = "Family History", 
              col_name = "Tumor Recurrence",
              placement = "top")

{janitor} tabyl with pct, without NAs

Missing data, or NAs, are included as a category by default. This is a good thing, so that you don’t miss the presence of missing data. But it is problematic if you are trying to do a chi square or fisher exact test. You can just make the tabyl argument show_na false. See the code chunk below for an example.

blood %>% 
  tabyl(gleason, tumor_vol, show_na = F)

Adding chi squared test to a janitor tabyl

Once the NA values are removed, you can then do a chi squared test. Just add one line, as in the code chunk below.
If you have a lot of low-count (<5) cells, you can also use a fisher exact test with fisher.test()

blood %>% 
  tabyl(gleason, tumor_vol, show_na = F) %>% 
  chisq.test()

## 
##  Pearson's Chi-squared test
## 
## data:  .
## X-squared = 43.318, df = 4, p-value = 8.888e-09

Your turn with supra!

Now try this with the supraclavicular data, with a table of group and gender.
Add a chi-squared test.

supra 

supra %>% 
  tabyl(group, gender)

supra %>% 
  tabyl(group, gender) %>% 
  chisq.test()

supra %>% 
  tabyl(group, gender) %>% 
  gt() %>% 
  tab_header(
    title = "Counts by Group",
    subtitle = "From the Supraclavicular Dataset"
  ) %>% 
  tab_options(
   table.background.color = "goldenrod",
   heading.background.color = "orange"
  ) 

supra %>% 
  tabyl(group, gender) %>% 
  untabyl() %>% 
  flextable() %>% 
  autofit() %>% 
  bold(part = "header") %>% 
  fontsize(part = "header", size = 20) %>% 
  fontsize(part = "body", size = 18) %>% 
  align(align = "center", part = "all") %>% 
  # note conditional formatting in rows when n>7, turn column n purple and bold. Could do this for P <0.05, etc.
  # rows denoted by i, columns denoted by j (as in i,j)
  color(color = "green", i =~ male>20, j =~ male) %>% 
  bold(i =~ male>20, j =~ male)

More Formatting

Note that if you want to use the formatting functions of gt or flextable, these are also options. Pipe your result into the functions

• as_gt() OR
• as_flextable()

Note - one possible gotcha - if you load the package flextable after the package gtsummary, you will need to specify which version of as_flextable() you want, as each has a function by that name. The easiest way to do this is to use gtsummary::as_flextable() so that R knows which version of the function that you want to use.

Your Turn with blood!

Now try gtsummary on the blood data set, using the variables age, aa, tumor_vol, and recur, divided by rbc_age categories, in the exercise below (Note - don’t eliminate rbc_age with your select statement)

blood 

blood %>% 
  select(Age, aa, tumor_vol, rbc_age, recur)

blood %>% 
  select(Age, aa, tumor_vol, rbc_age, recur) %>% 
  tbl_summary(by = rbc_age) %>% 
  bold_labels() %>%
  italicize_levels() %>%
  add_p() 

Adding p values for Comparisons with {gtsummary}

You can add comparative P values easily, see the chunk below (and its output) for an example.

supra %>% 
  select(gender, bmi, age, group) %>% 
  tbl_summary(by = group) %>% 
  add_p()

Your turn!

Now try this yourself on the blood dataset. Add p values for this comparison.
Which predictor variables do you think will be significant?

blood %>% 
  select(Age, aa, FamHx, recur, rbc_age) %>% 
  tbl_summary(by = recur,
  label = list(aa ~ "African-American",
                FamHx ~ "Family History of Prostate Ca",
               rbc_age ~ "Age of Transfused RBC"))

blood %>% 
  select(Age, aa, FamHx, recur, rbc_age) %>% 
  tbl_summary(by = recur,
  label = list(aa ~ "African-American",
                FamHx ~ "Family History of Prostate Ca",
               rbc_age ~ "Age of Transfused RBC")) %>% 
  add_p() %>% 
  bold_p(t = 0.10)

Your turn with licorice!

Now try this yourself on the licorice dataset. Add p values for this comparison.
Which predictor variables do you think will be significant?

licorice  %>% 
  select(gender, smoking, mallampati, treat, preOp_age) %>% 
  tbl_summary(by = treat)

licorice  %>% 
  select(gender, smoking, mallampati, treat, preOp_age) %>% 
  tbl_summary(by = treat) %>% 
  add_p() %>% 
  bold_p(t = 0.10)

Your turn!

Try to summarize in a table this model (in the blood dataset) for recurrence of prostate cancer with tbl_summary from the {gtsummary} package. Be sure to add the exponentiate = TRUE option to get odds ratios when you use tbl_summary()

model_recur_tbl <- glm(Recurrence ~ aa + fam_hx + rbc_age + tumor_vol + gleason,
          data = blood)

model_recur_tbl

model_recur_tbl <- glm(Recurrence ~ aa + fam_hx + rbc_age + tumor_vol + gleason,
                   data = blood) %>% 
   tbl_regression() 


model_recur_tbl

model_recur_tbl <- glm(Recurrence ~ aa + fam_hx + rbc_age + tumor_vol + gleason,
                   data = blood) %>% 
   tbl_regression(exponentiate = TRUE) 


model_recur_tbl

Quick Summary Table with gtsummary

First, this code chunk makes the summary table.

summ_tbl1 <- licorice %>% 
  select(gender, preOp_age, treat) %>% 
  tbl_summary(missing = "no") %>% 
  add_n()
summ_tbl1

Regression table with gtsummary

Then this code chunk builds the model, mod_tbl2,
and the tbl_regression() function makes the regression table

mod_tbl2 <- lm(pacu30min_throatPain ~ gender + smoking + mallampati + treat, data = licorice) %>% 
  tbl_regression()
mod_tbl2

Now merge these tables

Then you can merge these and add two tab spanners.

tbl_merge(
  list(summ_tbl1, mod_tbl2),
  tab_spanner = c("**Summary Statistics**",
                  "**30 Min Cough Outcome**")
)

summ_tbl_supra


regress_tbl_supra


merged_tbl_blood

summ_tbl_supra <- supra %>% 
  select(group, gender, bmi, age)

regress_tbl_supra <- lm(onset_motor ~ group + gender + bmi + age +fentanyl, data = supra)


merged_tbl_blood <-  tbl_merge(list(summ_tbl_supra, regress_tbl_supra))

summ_tbl_supra <- supra %>% 
  select(group, gender, bmi, age) %>% 
  tbl_summary() %>% 
  add_n()
summ_tbl_supra


regress_tbl_supra <- lm(onset_motor ~ group + gender + bmi + age +fentanyl, data = supra) %>% 
  tbl_regression()
regress_tbl_supra

merged_tbl_blood <-  tbl_merge(list(summ_tbl_supra, regress_tbl_supra),
            tab_spanner = c("**Summary Statistics**",
                  "**Onset Motor Outcome**"))
merged_tbl_blood