server.R

# Shiny App for visualization

function(input, output) {
  output$dens_ingrds <- renderPlot({
    if (input$orth) {
      if (input$samplesplit) {
        if (input$hq == "Tuned") {
          est_selection <- c("theta_dml")
        } else if (input$hq == "Underfitting") {
          est_selection <- c("theta_dml_underfit")
        } else if (input$hq == "Overfitting") {
          est_selection <- c("theta_dml_overfit")
        }
      } else {
        if (input$hq == "Tuned") {
          est_selection <- c("theta_orth_nosplit")
        } else if (input$hq == "Underfitting") {
          est_selection <- c("theta_dml_nosplit_underfit")
        } else if (input$hq == "Overfitting") {
          est_selection <- c("theta_dml_nosplit_overfit")
        }
      }
    } else {
      if (input$samplesplit) {
        if (input$hq == "Tuned") {
          est_selection <- c("theta_nonorth")
        } else if (input$hq == "Underfitting") {
          est_selection <- c("theta_nonorth_underfit")
        } else if (input$hq == "Overfitting") {
          est_selection <- c("theta_nonorth_overfit")
        }
      } else {
        if (input$hq == "Tuned") {
          est_selection <- c("theta_nonorth_nosplit")
        } else if (input$hq == "Underfitting") {
          est_selection <- c("theta_nonorth_nosplit_underfit")
        } else if (input$hq == "Overfitting") {
          est_selection <- c("theta_nonorth_nosplit_overfit")
        }
      }
    }

    data <- estimates_scaled[estimation == est_selection, ]

    g_dens <- ggplot(data) +
      geom_histogram(aes(
        y = ..density.., x = coef, fill = label,
        col = label
      ),
      bins = 30, alpha = 0.6
      ) +
      geom_vline(aes(xintercept = 0), col = "black") +
      suppressWarnings(geom_function(fun = dnorm, colour = "darkgrey")) +
      scale_color_manual(
        name = "",
        values = colors
      ) +
      scale_fill_manual(
        name = "",
        values = colors
      ) +
      xlim(c(-6.0, 6.0)) +
      xlab("") +
      ylab("") +
      theme_minimal() +
      theme(legend.position = "none")

    g_dens
  })

  output$dens_cf <- renderPlot({
    if (!input$crossfit) {
      est_selection <- c("theta_dml_nocf")
    } else {
      est_selection <- c("theta_dml")
    }

    data <- estimates_unscaled[estimation == est_selection, ]

    g_dens_unscaled <- ggplot(data) +
      geom_histogram(aes(
        y = ..density.., x = coef, fill = label,
        col = label
      ),
      bins = 30, alpha = 0.9
      ) +
      geom_vline(aes(xintercept = 0), col = "black") +
      # suppressWarnings(geom_function(fun = dnorm, colour = "darkgrey")) +
      scale_color_manual(
        name = "",
        values = colors
      ) +
      scale_fill_manual(
        name = "",
        values = colors
      ) +
      xlim(c(-0.25, 0.25)) +
      ylim(c(0, 9)) +
      xlab("") +
      ylab("") +
      theme_minimal() +
      theme(legend.position = "none")

    g_dens_unscaled
  })
}