qwraps2: Summary Table

In qwraps2: Quick Wraps 2

knitr::opts_chunk$set(collapse = TRUE)

set.seed(42)
library(qwraps2)
options(qwraps2_markup = "markdown")

The r backtick(summary_table) method appears to be the most popular and widely used feature of the r CRANpkg(qwraps2) package. As such, this vignette is provided to give as much detail on the use of the method, and the underlying r backtick(qable) method for quickly building well formatted summary tables.

qable

r backtick(qable) builds a formatted character matrix from inputs and then renders a table via knitr::kable. The primary objective of this function is to allow for easy construction of row groups.

kable vs qable

For a simple example we will use the following data set with a grouping variable, subject id, and two variables, V2, and V3. For simplicity, we will order the data by group and id as well.

d <- data.frame(
       group = sample(size = 15, paste0("grp", 1:5), replace = TRUE)
     , id = sample(size = 15, x = LETTERS)
     , V2 = rnorm(15)
     , V3 = rep(c(1, 2, NA), times = 5)
     )
d <- d[order(d$group, d$id), ]

Making a simple table via kable:

knitr::kable(d, row.names = FALSE)

The group column is great for data analysis, but is not the best for human readability. This is where r backtick(qable) can be useful. Start by building a named numeric column with the name being the row group name and the value the number of rows. For the ordered data this is a simple call to table:

c(table(d$group))

If we pass that named vector to r backtick(qable) as the rgroup and with specify the id column as the row names we have the same information but in format that is better for humans:

qable(  x = d[, c("V2", "V3")]
      , rgroup = c(table(d$group)) # row group
      , rnames = d$id              # row names
)

The return object from r backtick(qable) is a character matrix. Also, when a data.frame is passed to r backtick(qable) it is coerced to a matrix before anything else, as such, any formatting of numeric values or other strings should be done before calling r backtick(qable) %s% "."

To pass arguments to knitr::kable do so via the r backtick(kable_args) argument.

Example: Regression Model Summary Table

We will build a summary table for a regression model with row groups for conceptually similar predictors.

model <-
  glm(spam ~
        word_freq_your + word_freq_conference + word_freq_business +
        char_freq_semicolon + char_freq_exclamation_point +
        capital_run_length_total + capital_run_length_longest
    , data = spambase
    , family = binomial()
  )

model_summary <-
  data.frame(
    parameter = names(coef(model))
  , odd_ratio = frmt(exp(coef(model)), digits = 3)
  , lcl       = frmt(exp(coef(model) + qnorm(0.025) * sqrt(diag(vcov(model)))), digits = 3)
  , ucl       = frmt(exp(coef(model) + qnorm(0.975) * sqrt(diag(vcov(model)))), digits = 3)
  , pval      = frmtp(summary(model)$coef[, 4])
  )

qable(model_summary[-1, c('odd_ratio', 'lcl', 'ucl', 'pval')]
      , rtitle = "Parameter"
      , rgroup = c("Word Frequency" = 3, "Character Frequency" = 2, "Capital Run Length" = 2)
      , rnames = c("Your", "Conference", "Business", ";", "!", "Total", "Longest")
      , kable_args = list(align = "lrrrr", caption = "Regression Model Summary")
      , cnames = c("Odds Ratio", "Lower Conf. Limit", "Upper Conf. Limit", "P-value")
      )

summary_table

r backtick(summary_table) was developed with the primary objective to build well formatted and easy to read data summary tables. Conceptually, the construction of these tables start by building a "list-of-lists" of summaries and then generating these summaries for specific groupings of the data set.

Defining a Summary

We will use the r backtick(mtcars2) data set for these examples. We'll start with something very simple and build up to something bigger.

Let's report the min, max, and mean (sd) for continuous variables and n (%) for categorical variables. We will report mpg, displacement (disp), wt (weight), and gear overall and by number of cylinders and transmission type.

The use of the r backtick(summary_table) use to define a summary, that is, a list-of-lists of formulas for summarizing the data.frame.

The inner lists are named formulae defining the wanted summary. The names are important, as they are used to label row groups and row names in the table.

our_summary1 <-
  list("Miles Per Gallon" =
       list("min"       = ~ min(mpg),
            "max"       = ~ max(mpg),
            "mean (sd)" = ~ qwraps2::mean_sd(mpg)),
       "Displacement" =
       list("min"       = ~ min(disp),
            "median"    = ~ median(disp),
            "max"       = ~ max(disp),
            "mean (sd)" = ~ qwraps2::mean_sd(disp)),
       "Weight (1000 lbs)" =
       list("min"       = ~ min(wt),
            "max"       = ~ max(wt),
            "mean (sd)" = ~ qwraps2::mean_sd(wt)),
       "Forward Gears" =
       list("Three" = ~ qwraps2::n_perc0(gear == 3),
            "Four"  = ~ qwraps2::n_perc0(gear == 4),
            "Five"  = ~ qwraps2::n_perc0(gear == 5))
       )

Building the table is done with a call to r backtick(summary_table) and rendered in Table \@ref(tab:mtcars_whole).

whole <-
  summary_table(
    x = mtcars2
  , summaries = our_summary1
  , qable_args = list(kable_args = list(caption = "mtcars2 data summary"))
  )
whole

Summarize by

Use the r backtick(by) argument to specify a grouping variable and generate the same summary as above but for subsets of the data. When the r backtick(by) column is a factor, the columns will be in the order of the levels of the factor. In comparison, the column order is alphabetical if the variable is just a character.

by_cylf <-
  summary_table(
    x = mtcars2
  , summaries = our_summary1
  , by = c("cyl_factor")
  , qable_args = list(rtitle = "Summary Statistics"
                      , kable_args = list(caption = "mtcars2 data summary by cyl_factor"))
  )
by_cylf

by_cylc <-
  summary_table(
    x = mtcars2
  , summaries = our_summary1
  , by = c("cyl_character")
  , qable_args = list(rtitle = "Summary Statistics"
                      , kable_args = list(caption = "mtcars2 data summary by cyl_character"))
  )
by_cylc

You are also able to generate summaries by multiple columns. For example, Table \@ref(tab:mtcars2_by_cyl_transmission) reports the summary by the combination of the number of cylinders and the type of transmission.

by_cyl_am <-
  summary_table(
    x = mtcars2
  , summaries = our_summary1
  , by = c("cyl_factor", "transmission")
  )
by_cyl_am

cbind summary_table

It is common that I will want to have a summary table with the first column reporting for the whole data sets and the additional columns for subsets of the data set. The returned objects from r backtick(summary_table) can be joined together via r backtick(cbind) assuming that the row groupings (summaries) are the same.

Note: the r backtick(kable_args) of the first item passed to r backtick(cbind) will be assigned to the resulting object (Table \@ref(tab:mtcars2_cbind)). However, there is an easy way to modify the qable_args and kable_args via the print method.

both <- cbind(whole, by_cylf)
both

If you want to update how a summary table is printed, you can do so by calling the print method explicitly while passing a new set of r backtick(qable_args) %s% "," see Table \@ref(tab:updated_both).

print(both,
      qable_args = list(
        rtitle = "ROW-TITLE",
        cnames = c("Col 0", "Col 1", "Col 2", "Col 3"),
        kable_args = list(
          align = "lcrcr",
          caption = "mtcars2 data summary - new caption"
        )
      ))

Adding P-values to a Summary Table

There are many different ways to format data summary tables. Adding p-values to a table is just one thing that can be done in more than one way. For example, if a row group reports the counts and percentages for each level of a categorical variable across multiple (column) groups, then I would argue that the p-value resulting from a chi square test or a Fisher exact test would be best placed on the line of the table labeling the row group. However, say we reported the minimum, median, mean, and maximum with in a row group for one variable. The p-value from a t-test, or other meaningful test for the difference in mean, I would suggest should be reported on the line of the summary table for the mean, not the row group itself.

With so many possibilities I have reserved construction of a p-value column to be ad hoc. Perhaps an additional column wouldn't be used and the p-values are edited into row group labels, for example.

If you want to add a p-value column, or any other column(s) to a r backtick(qwraps2_summary_table) object you can with some degree of ease. Note that r backtick(qwraps2_summary_table) objects are just character matrices with additional attributes.

str(both)

For this example, we will added p-values for testing the difference in the mean between the three cylinder groups and the distribution of forward gears by cylinder groups.

# difference in means
mpvals <-
  sapply(
         list(mpg  = lm(mpg  ~ cyl_factor, data = mtcars2),
              disp = lm(disp ~ cyl_factor, data = mtcars2),
              wt   = lm(wt   ~ cyl_factor, data = mtcars2)),
         extract_fpvalue)

# Fisher test
fpval <- frmtp(fisher.test(table(mtcars2$gear, mtcars2$cyl_factor))$p.value)

In this case, adding the p-value column, is done by creating a empty column and then writing in the needed p-value on the wanted rows. This could be within a row group (tests for means) or for a row group (Fisher test).

both <- cbind(both, "P-value" = "")
both[grepl("mean \\(sd\\)", both[, 1]), "P-value"] <- mpvals
both[grepl("Forward Gears", both[, 1]), "P-value"] <- fpval

print(both, qable_args = list(kable_args = list(caption = "mtcars2 summary with p-values")))

Another option you might consider is to have the p-value in the row group name. Consider the following construction. The p-values are added to the names of the row groups when building the summary table.

gear_summary <-
  list("Forward Gears" =
       list("Three" = ~ qwraps2::n_perc0(gear == 3),
            "Four"  = ~ qwraps2::n_perc0(gear == 4),
            "Five"  = ~ qwraps2::n_perc0(gear == 5)),
       "Transmission" =
       list("Automatic" = ~ qwraps2::n_perc0(am == 0),
            "Manual"    = ~ qwraps2::n_perc0(am == 1))
       )

gear_summary <-
setNames(gear_summary,
         c(
         paste("Forward Gears: ", frmtp(fisher.test(xtabs( ~ gear + cyl_factor, data = mtcars2))$p.value)),
         paste("Transmission: ",  frmtp(fisher.test(xtabs( ~ am + cyl_factor, data = mtcars2))$p.value)))
         )

summary_table(mtcars2, gear_summary, by = "cyl_factor")

rbind summary_table

There is a rbind method of summary tables. This can be useful when building a large a table in smaller sections would be advantageous. For example, it might be helpful to add p-values to a summary table with just one row group and then rbind all the tables together for printing. Consider that in the above example for adding p-values we have made an assumption that the order of the summary and the r backtick(mpvals) will be static. Remembering to make the sequence changes in more than one location can be more difficult than we would like to admit. Writing code to be robust to such changes is preferable.

t_mpg  <- summary_table(mtcars2, summaries = our_summary1["Miles Per Gallon"], by = "cyl_factor")
t_disp <- summary_table(mtcars2, summaries = our_summary1["Displacement"], by = "cyl_factor")
t_wt   <- summary_table(mtcars2, summaries = our_summary1["Weight (1000 lbs)"], by = "cyl_factor")

t_mpg  <- cbind(t_mpg,  "pvalue" = "")
t_disp <- cbind(t_disp, "pvalue" = "")
t_wt   <- cbind(t_wt,   "pvalue" = "")

t_mpg[ grepl("mean", t_mpg[, 1]),  "pvalue"] <- "mpg-pvalue"
t_disp[grepl("mean", t_disp[, 1]), "pvalue"] <- "disp-pvalue"
t_wt[  grepl("mean", t_wt[, 1]),   "pvalue"] <- "wt-pvalue"

Calling rbind now will let us have the table in different sequences without having to worry about the alignment of rows between different elements:

rbind(t_mpg, t_disp, t_wt)
rbind(t_wt, t_disp, t_mpg)

Using Variable Labels

Some data management paradigms will use attributes to keep a label associated with a variable in a data.frame. Notable examples are the r CRANpkg(Hmisc) and r CRANpkg(sjPlot) %s% "." If you associate a label with a variable in the data frame the that label will be used when building a summary table. This feature was suggested https://github.com/dewittpe/qwraps2/issues/74 and implemented thusly:

new_data_frame <-
  data.frame(age = c(18, 20, 24, 17, 43),
             edu = c(1, 3, 1, 5, 2),
             rt  = c(0.01, 0.04, 0.02, 0.10, 0.06))

# Set a label for the variables
attr(new_data_frame$age, "label") <- "Age in years"
attr(new_data_frame$rt,  "label") <- "Reaction time"

# mistakenly set the attribute to name instead of label
attr(new_data_frame$edu, "name") <- "Education"

When calling r backtick(qsummary) the provide labels for the age and rt variables will be used. Since the attribute "label" does not exist for the edu variable, edu will be used in the output.

qsummary(new_data_frame)

This behavior is also seen with the r backtick(summary_table) call.

summary_table(new_data_frame)

Alternative building of the summaries

The task of building the r backtick(summaries) list-of-lists can be tedious. The function r backtick(qummaries) is designed to make it easier. r backtick(qummaries) will use a set of predefined functions to summarize numeric columns of a data.frame, a set of arguments to pass to r backtick(n_perc) for categorical (character and factor) variables.

By default, calling r backtick(summary_table) will use the default summary metrics defined by r paste0(backtick(qsummary), ".") The purpose of r backtick(qsummary) is to provide the same summary for all numeric variables within a data.frame and a single style of summary for categorical variables within the data.frame. For example, the default summary for a set of variables from the r backtick(mtcars2) data set is

qsummary(mtcars2[, c("mpg", "cyl_factor", "wt")])

That default summary is used for a table as follows:

summary_table(mtcars2[, c("mpg", "cyl_factor", "wt")])

Now, say we want to only report the minimum and maximum for each of the numeric variables and for the categorical variables we want two show the denominator for each category and for the percentage, to one digit with the percent symbol in the table. Note that when defining the list of numeric_summaries that the argument place holder is the r backtick("%s%", dequote = TRUE) character.

new_summary <-
  qsummary(mtcars2[, c("mpg", "cyl_factor", "wt")],
           numeric_summaries = list("Minimum" = "~ min(%s)",
                                    "Maximum" = "~ max(%s)"),
           n_perc_args = list(digits = 1, show_symbol = TRUE, show_denom = "always"))
str(new_summary)

The resulting table is:

summary_table(mtcars2, new_summary)

Session Info

print(sessionInfo(), local = FALSE)

qwraps2 documentation built on Nov. 10, 2023, 1:06 a.m.