Merging column header in latex table created by kable in Rmarkdown

Question

I have created a latex table as below using kable in Rmarkdown:

---
output: pdf_document
header-includes:
  - \usepackage{xcolor}
---

```{r, message=FALSE, warning=FALSE, echo=FALSE}
library(kableExtra)
library(tidyr)
library(dplyr)

data(iris)
iris %>% 
  as_tibble %>%
  gather(.,key = variable,value = value,-Species) %>%
  group_by(Species,variable) %>%
  summarise(value=mean(value)) %>%
  ungroup %>%
  spread(.,key = variable,value = value) %>%
  mutate(`Percentage Change`=`Petal.Length`/`Petal.Width`*100) %>%
        kable(.,format='latex',
        align='c',linesep='',
        booktabs=TRUE,escape=FALSE) %>%
  add_header_above(.,c(' '=1,'Parts'=4,' '=1),
                   escape = FALSE) %>%
  kable_styling(latex_options = c('striped','HOLD_position','scale_down'))
```

I would like to have the column header "Species" and "Percentage Change" merged with the empty space above them respectively, so that Species can be placed in the middle of the two header rows, while Percentage Change (Petal Length/ Petal Width) can occupy two rows, rather than having a empty row above, and prevent other column to have an empty row below.

Wonder if it can be modified in kable preferably, latex "hack" suggestion is also welcome.

Thanks!

Answer 1

I think for this latex 'hack' solution is much cleaner. In kable also this can be done but that would require changing the data frame (convert column names to row) so that collapse_rows can be used. Anyway, here's the latex way out:

The code that you gave in your questions does not give the column name as in the pdf snapshot. So I edited the code first to get that table:

---
output: 
  pdf_document:
    keep_tex: true
header-includes:
  - \usepackage{xcolor}
---

```{r, message=FALSE, warning=FALSE, echo=FALSE}
library(kableExtra)
library(tidyr)
library(dplyr)

data(iris)
iris %>% 
  as_tibble %>%
  gather(.,key = variable,value = value,-Species) %>%
  group_by(Species,variable) %>%
  summarise(value=mean(value)) %>%
  ungroup %>%
  spread(.,key = variable,value = value) %>%
  mutate('Percentage Change\n(Petal length/ Petal width)'=`Petal.Length`/`Petal.Width`*100) %>%
  kable(format='latex',align='c',linesep='',booktabs=TRUE,escape=FALSE,
        col.names = linebreak(colnames(.),align = 'c')) %>%
  add_header_above(.,c(' '=1,'Parts'=4,' '=1),escape = FALSE) %>%
  collapse_rows(columns = c(1,6),valign = 'middle')%>%
  kable_styling(latex_options = c('striped','HOLD_position','scale_down'))
```

This gives this:

Note two things in above code:

1. keep_tex: true: this retains the .tex file generated and can be used to edit.
2. Use of linebreaks to ensure that the entire column name for last column is not in one line.

Now we make small changes in latex output. In code below the commented out line is the original code generated by kable. This is replaced by the new lines just below the commented out line as indicated.

\begin{table}[H]
\centering
\resizebox{\linewidth}{!}{
\begin{tabular}{cccccc}
\toprule
% \multicolumn{1}{c}{ } & \multicolumn{4}{c}{Parts} & \multicolumn{1}{c}{ } \\
\multirow{2}{*}{Species} & \multicolumn{4}{c}{Parts} & \multirow{2}{*}{\makecell[c]{Percentage Change\\(Petal length/ Petal width)}} \\ % replaced line
\cmidrule(l{3pt}r{3pt}){2-5}
% Species & Petal.Length & Petal.Width & Sepal.Length & Sepal.Width & \makecell[c]{Percentage Change\\(Petal length/ Petal width)}\\
& Petal.Length & Petal.Width & Sepal.Length & Sepal.Width &\\ % replaced line
\midrule
\cellcolor{gray!6}{setosa} & \cellcolor{gray!6}{1.462} & \cellcolor{gray!6}{0.246} & \cellcolor{gray!6}{5.006} & \cellcolor{gray!6}{3.428} & \cellcolor{gray!6}{594.3089}\\
\cmidrule{1-6}
versicolor & 4.260 & 1.326 & 5.936 & 2.770 & 321.2670\\
\cmidrule{1-6}
\cellcolor{gray!6}{virginica} & \cellcolor{gray!6}{5.552} & \cellcolor{gray!6}{2.026} & \cellcolor{gray!6}{6.588} & \cellcolor{gray!6}{2.974} & \cellcolor{gray!6}{274.0375}\\
\bottomrule
\end{tabular}}
\end{table}

This gives the following output: