Summary Tables with ‘tab’

Functions

The purpose of tab is to make it easier to create tables for papers, including Table 1’s showing characteristics of the sample and summary tables for fitted regression models. Currently, the following functions are included:

• tabmeans compares means in two or more groups.
• tabmedians compares medians in two or more groups.
• tabfreq compares frequencies in two or more groups.
• tabmulti compares multiple variables in two or more groups.
• tabmeans.svy, tabmedians.svy, tabfreq.svy, and tabmulti.svy serve the same purposes as the above functions, but for complex survey data.
• tabglm summarizes generalized linear models (GLM’s) fit via glm or survey::svyglm.
• tabgee summarizes generalized estimating equation models (GEE’s) fit via gee::gee.
• tabcoxph summarizes Cox Proportional Hazards models fit via survival::coxph or survey::svycoxph.

Table 1’s

You can use tabmulti to compare characteristics across levels of a factor variable, e.g. here comparing age, sex, and race by treatment group in the toy dataset tabdata.

tabmulti(Age + Sex + Race ~ Group, data = tabdata) %>% kable()

To illustrate some options, we can request Age and Race to print as Age (years) and Race/ethnicity, compare medians rather than means for age, and include the sample sizes in the column headings:

tabmulti(Age + Sex + Race ~ Group, data = tabdata, 
         yvarlabels = list(Age = "Age (years)", Race = "Race/ethnicity"), 
         ymeasures = c("median", "freq", "freq"), 
         listwise.deletion = TRUE, 
         n.headings = TRUE) %>% kable()

Regression tables

fit <- glm(death_1yr ~ Age + Sex + Group, data = tabdata, family = binomial)
fit %>% tabglm(factor.compression = 5) %>% kable()

tabdata2 <- reshape(data = tabdata,
                    varying = c("bp.1", "bp.2", "bp.3", "highbp.1", "highbp.2", "highbp.3"),
                    timevar = "bp.visit", direction = "long")
tabdata2 <- tabdata2[order(tabdata2$id), ]
fit <- gee(highbp ~ poly(Age, 2, raw = TRUE) + Sex + Race + Race*Sex,
           id = id, data = tabdata2, family = "binomial", corstr = "unstructured")
fit %>% tabgee(data = tabdata2) %>% kable()

library("survival")
fit <- coxph(Surv(time = time, event = delta) ~ Age + Sex + Group, data = tabdata)
fit %>% tabcoxph(factor.compression = 5, columns = c("beta", "hr.ci")) %>% kable()

Complex survey data

The functions in tab can also accommodate complex survey data. To illustrate with the included dataset tabsvydata (which is data from NHANES 2003-2004, except for the made-up variables time and event), here’s a Table 1:

library("survey")
design <- svydesign(
  data = tabsvydata,
  ids = ~sdmvpsu,
  strata = ~sdmvstra,
  weights = ~wtmec2yr,
  nest = TRUE
)
tabmulti.svy(Age + Race + BMI ~ Sex, design = design) %>% kable()

And here’s a linear regression:

fit <- svyglm(BMI ~ Age + Sex + Race, design = design)
fit %>% tabglm(factor.compression = 3) %>% kable()

Exporting tables, e.g. to Word

All of the functions in tab have an argument called print.html which can be used to export tables to word processors. Setting print.html = TRUE will result in a HTML table being output to your current working directory. You can open the table (e.g. in Chrome) and copy/paste into your report.

Options for printing in R

I used knitr’s kable function for the examples here, but other approaches should also work (e.g. xtable’s xtable or pandoc’s pandoc.table).