r - 汇总由 R knitr 中的模板动态生成的结果

Question

我正在使用knitr, 使用模板编写报告，该模板对多个数据集进行分析（MWE 的第 1 节）。

我可以通过将值分配给变量来生成结果摘要，然后将这些变量“拼接”到表格中（MWE 的第 2 节）。然而，这种方法既麻烦又不灵活（例如，为了改变表中出现的特定位需要大量输入）。

如何自动生成汇总表？

• 报告正文 (MWE.Rnw)：

  \documentclass{article}
  
  \begin{document}
  
  \tableofcontents
  
  \newpage
  \section{Run tests}
  
  In this section we run the tests using a template.
  
  <<run-all, include = FALSE>>=
  library(knitr)
  
  ## set data names
  data_names <- LETTERS[1:3]
  
  ## initialize var for data
  data_1 <- NULL
  data_2 <- NULL
  
  ## initialize vars for chi-squared test results
  cs_statistic <- NULL                    # X-squared
  cs_parameter <- NULL                    # df
  cs_p_value <- NULL                      # p-value
  
  ## initialize vars for binomial test results
  bt_p_value <- NULL                      # p-value
  bt_estimate <- NULL                     # estimate
  bt_ci_lower <- NULL                     # conf. int. lower
  bt_ci_upper <- NULL                     # conf. int. upper
  
  ## run template
  src = NULL
  for (i in data_names) src = c(src, knit_expand('analysis-template.Rnw'))
  @
  
  \Sexpr{paste(knit(text = src), collapse = '\n')}
  
  \newpage
  \section{Summary}
  
  In this section we summarise the results.
  
  <<summary-cs>>=
  tab <- data.frame(data_1, data_2, round(cs_statistic, 3), cs_parameter, round(cs_p_value, 3), row.names = data_names)
  colnames(tab) <- c("var 1", "var 2", "X-squared", "d.f.", "p-value")
  
  kable(tab, caption = "Summary results of $\\chi^2$ tests")
  @ 
  
  <<summary-bt>>=
  tab <- data.frame(data_1, data_2, round(bt_estimate, 3), round(bt_ci_lower, 3), round(bt_ci_upper, 3), round(bt_p_value, 3), row.names = data_names)
  colnames(tab) <- c("var 1", "var 2", "estimate", "95% conf. int. (lower)", "95% conf. int. (upper)", "p-value")
  
  kable(tab, caption = "Summary results of binomial tests")
  @ 
  
  \end{document}
  

• MWE.Rnw调用的模板（analysis-template.Rnw）：

  \subsection{Analysis of data {{i}}}
  
  <<analysis-{{i}}>>=
  ## generate data {{i}}
  (data_{{i}} <- sample(50:100, 2))
  
  ## run tests
  (cs <- chisq.test(data_{{i}}))
  (bt <- binom.test(data_{{i}}[1], sum(data_{{i}})))
  
  ## store results
  data_1 <- c(data_1, data_{{i}}[1])
  data_2 <- c(data_2, data_{{i}}[2])
  
  cs_statistic <- c(cs_statistic, cs$statistic)
  cs_parameter <- c(cs_parameter, cs$parameter)
  cs_p_value <- c(cs_p_value, cs$p.value)
  
  bt_estimate <- c(bt_estimate, bt$estimate)
  bt_ci_lower <- c(bt_ci_lower, bt$conf.int[1])
  bt_ci_upper <- c(bt_ci_upper, bt$conf.int[2])
  bt_p_value <- c(bt_p_value, bt$p.value)
  @