1

我刚开始尝试使用Shiny，但让dplyr 'group_by' 函数工作时遇到了问题。

我的一般想法是，我有一个包含许多数值变量的数据集和一个可以对这些变量进行分组的变量（在我的例子中是 19 个国家，所以变量范围从 1 到 19）。现在我想使用metafor包对每个国家的两个变量之间的相关性进行荟萃分析。所以我想分别计算每个国家的两个选定变量之间的相关性，然后将它们放在一起进行荟萃分析，最后显示这些影响的森林图。

我以 mtcars 数据集为例说明我已经走了多远：

用户界面：

library(shiny)
library(metafor)
library(dplyr)


ui <- fluidPage(
  sidebarPanel(

    selectInput("x", label = "Choose Variable I", 
    choices = c("Displacement (cu.in.)", "Horsepower", "Rear axle ratio",
    "Weight (1000lbs)", "1/4 mile time"), selected = "Displacement (cu.in.)"),

    selectInput("y", label = "Choose Variable II", 
    choices = c("Displacement (cu.in.)", "Horsepower", "Rear axle ratio",
    "Weight (1000lbs)", "1/4 mile time"), selected = "Horsepower")
  ),

  mainPanel(
    plotOutput(outputId = "plot")
    ) 
  )

服务器：

 server <- function(input, output) {


  output$plot <-renderPlot({

    data_x <- switch(input$x, 
                     "Displacement (cu.in.)" = mtcars$disp,
                     "Horsepower" = mtcars$hp,
                     "Rear axle ratio" = mtcars$drat,
                     "Weight (1000lbs)" = mtcars$wt,
                     "1/4 mile time" = mtcars$qsec)

    data_y <- switch(input$y, 
                     "Displacement (cu.in.)" = mtcars$disp,
                     "Horsepower" = mtcars$hp,
                     "Rear axle ratio" = mtcars$drat,
                     "Weight (1000lbs)" = mtcars$wt,
                     "1/4 mile time" = mtcars$qsec)

    meta_main <- mtcars %>% 
      group_by(gear) %>% 
      summarise(participantID = n(), 
                correlation = cor(data_x, data_y, use = "complete.obs"))


    meta <- rma(ni=participantID, ri=correlation, 
                method="REML", measure="COR", data=meta_main)


    forest(meta) 

  })

}

shinyApp(ui = ui, server = server)

在示例中，我按“齿轮”（有 3 个级别）进行分组，并选择了变量“排量”、“马力”、“后轴比”、“重量”和“1/4 英里时间”。

最终输出显示选择小部件以及森林图。但是，所有级别的“齿轮”的相关性都是相同的。因此，我怀疑“group_by”功能无法按预期工作。

我尝试按照网上的建议使用“group_by_”，但变化不大。

关于“group_by”如何与 Shiny 一起工作的任何想法？

4

1 回答 1

1

试试这个服务器功能，首先你应该加载lazyeval包。

server <- function(input, output) {

output$plot <-renderPlot({

data_x <- switch(input$x, 
                 "Displacement (cu.in.)" = "disp",
                 "Horsepower" = "hp",
                 "Rear axle ratio" = "drat",
                 "Weight (1000lbs)" = "wt",
                 "1/4 mile time" = "qsec")

data_y <- switch(input$y, 
                 "Displacement (cu.in.)" = "disp",
                 "Horsepower" = "hp",
                 "Rear axle ratio" = "drat",
                 "Weight (1000lbs)" = "wt",
                 "1/4 mile time" = "qsec")

meta_main <- mtcars %>% 
  group_by(gear) %>% 
  mutate(participantID = n()) %>%
  group_by(gear, participantID) %>% 
  summarise_(correlation = interp(~cor(x, y, use = "complete.obs"), x = as.name(data_x), y = as.name(data_y)))


meta <- rma(ni=participantID, ri=correlation, 
            method="REML", measure="COR", data=meta_main)


forest(meta) 

})

}
于 2017-12-04T10:25:32.207 回答