R/DescriptiveStrategy_m2.R

In LOGAN: Log File Analysis in International Large-Scale Assessments

#' Report: Descriptive statistics by strategy
#'
#' This is a function that reports a descriptive analysis of the strategy and
#' students performance
#'
#' @param data A \code{matrix} or \code{data.frame} where the 'strategy.var' and
#'   performance variables are
#' @param strategy.var A character string with the name of the strategy variable
#' @param performance.item A character string with the name of the item
#'   performance variable
#' @param performance.test A character string with the name of the test
#'   performance variable
#' @param PartialCredit Logical. It can be used when the item is partial credit
#'   score.
#'
#' @return This function returns a report with a descriptive analysis of the
#'   strategy and students performance
#'
#' @examples
#' m2$DescriptiveStrategy(cp025q01.treated, "votat", "CP025Q01", "PV1CPRO")
#'
DescriptiveStrategy <- function(data, strategy.var, performance.item,
                                performance.test, PartialCredit = FALSE) {
  if (PartialCredit == TRUE) {
    # Frequency table (N): categorical variables with partialCredit: only
    # accepts options: 0, 0.5 and 1
    tab.freqitem <- as.data.frame(table(data[, performance.item]))
    for (i in seq(length(tab.freqitem$Var1))) {
      if (i == 1 & tab.freqitem$Var1[i] != 0.5) {
        scalename2 <- tab.freqitem$Var1[i]
        n2 <- tab.freqitem$Freq[i]
        total2 <- tab.freqitem$Freq[i]
        perc2 <- sprintf("%.0f%%", 100 * n2 / sum(tab.freqitem$Freq))
      } else if (tab.freqitem$Var1[i] == 0.5) {
        i <- i + 1
        scalename2 <- paste0(scalename2, "|", tab.freqitem$Var1[i])
        n2 <- paste0(n2, "|", tab.freqitem$Freq[i - 1] + tab.freqitem$Freq[i])
        total2 <- total2 + tab.freqitem$Freq[i - 1] + tab.freqitem$Freq[i]
        perc2 <- paste0(
          perc2, "|",
          sprintf("%.0f%%", 100 *
            (tab.freqitem$Freq[i - 1] * 0.5 +
              tab.freqitem$Freq[i]) /
            sum(tab.freqitem$Freq))
        )
      }
    }
  } else {
    tab.freqitem <- as.data.frame(table(data[, performance.item]))
    for (i in seq(length(tab.freqitem$Var1))) {
      if (i == 1) {
        scalename2 <- tab.freqitem$Var1[i]
        n2 <- tab.freqitem$Freq[i]
        total2 <- tab.freqitem$Freq[i]
        perc2 <- sprintf("%.0f%%", 100 * n2 / sum(tab.freqitem$Freq))
      } else {
        scalename2 <- paste0(scalename2, "|", tab.freqitem$Var1[i])
        n2 <- paste0(n2, "|", tab.freqitem$Freq[i])
        total2 <- total2 + tab.freqitem$Freq[i]
        perc2 <- paste0(
          perc2, "|",
          sprintf("%.0f%%", 100 * tab.freqitem$Freq[i] /
            sum(tab.freqitem$Freq))
        )
      }
    }
  }
  # Frequency table (N): categorical variables
  tab.freqvotat <- as.data.frame(table(data[, strategy.var]))
  for (i in seq(length(tab.freqvotat$Var1))) {
    if (i == 1) {
      scalename1 <- tab.freqvotat$Var1[i]
      n1 <- tab.freqvotat$Freq[i]
      total1 <- tab.freqvotat$Freq[i]
      perc1 <- sprintf("%.0f%%", 100 * tab.freqvotat$Freq[i] / sum(tab.freqvotat$Freq))
    } else {
      scalename1 <- paste0(scalename1, "|", tab.freqvotat$Var1[i])
      n1 <- paste0(n1, "|", tab.freqvotat$Freq[i])
      total1 <- total1 + tab.freqvotat$Freq[i]
      perc1 <- paste0(
        perc1, "|",
        sprintf("%.0f%%", 100 * tab.freqvotat$Freq[i] /
          sum(tab.freqvotat$Freq))
      )
    }
  }

  tab.print <-
    as.data.frame(rbind(
      cbind(paste0("Strategy: ", strategy.var), scalename1, perc1, n1, total1),
      cbind(paste0("Performance: ", performance.item), scalename2, perc2, n2, total2)
    ))

  colnames(tab.print) <- c("Construct", "Scale", "Relat. Freq.", "Frequencies", "Total N")
  message("\n Frequency table - Individual level")
  pander::pandoc.table(tab.print, split.tables = 100)

  # Chi-squared test of independence
  crostab.freqvotat <- table(data[, c(strategy.var, performance.item)])
  xtest.crostab <- stats::chisq.test(crostab.freqvotat)

  message(
    "Measures of association between ", strategy.var, " and ",
    performance.item, " - Individual level"
  )
  crostab.freqvotat <- rbind(crostab.freqvotat, apply(crostab.freqvotat, 2, sum))
  crostab.freqvotat <- cbind(crostab.freqvotat, apply(crostab.freqvotat, 1, sum))
  crostab.freqvotat <- cbind(
    row.names(crostab.freqvotat),
    as.data.frame.matrix(crostab.freqvotat,
      row.names = seq(dim(crostab.freqvotat)[1])
    )
  )
  crostab.freqvotat[, 1] <- as.character(crostab.freqvotat[, 1])
  crostab.freqvotat[dim(crostab.freqvotat)[1], 1] <- "Total"
  names(crostab.freqvotat)[1] <- paste0(strategy.var, " / ", performance.item)
  names(crostab.freqvotat)[length(names(crostab.freqvotat))] <- "Total"

  value.p <- ifelse(round(xtest.crostab$p.value, 4) < 0.01, "0.01",
    round(xtest.crostab$p.value, 4)
  )
  pander::pandoc.table(crostab.freqvotat, split.tables = 100)
  message(paste0(
    "Chi-squared = ", round(xtest.crostab$statistic, 2), ", df = ",
    xtest.crostab$parameter, ", p-value < ", value.p, "\n"
  ))

  if (prod(dim(table(data[, c(strategy.var, performance.item)]))) == 4) {
    phi.crostab <- psych::phi(table(data[, c(strategy.var, performance.item)]))
    message(paste0("Phi coefficient = ", round(phi.crostab, 4), "\n"))
  }

  # Summary table: continua variables - Summary of  PV1CPRO  by strategy
  for (j in seq(length(tab.freqvotat$Var1))) {
    data2 <- as.data.frame(data[data[, strategy.var] == as.character(tab.freqvotat$Var1[j]), performance.test])
    names(data2) <- performance.test

    if (j == 1) {
      tab.perftest <- as.matrix.data.frame(rbind(
        round(length(data2[[performance.test]]), 0),
        round(min(data2[[performance.test]]), 2),
        round(stats::quantile(data2[[performance.test]],
          probs = 0.25
        ), 2),
        round(stats::median(data2[[performance.test]]), 2),
        round(mean(data2[[performance.test]]), 2),
        round(stats::sd(data2[[performance.test]]), 2),
        round(stats::quantile(data2[[performance.test]],
          probs = 0.75
        ), 2),
        round(max(data2[[performance.test]]), 2)
      ))
      colnames(tab.perftest)[j] <- as.character(tab.freqvotat$Var1[j])
    } else {
      tab.perftest2 <- as.matrix.data.frame(rbind(
        round(length(data2[[performance.test]]), 0),
        round(min(data2[[performance.test]]), 2),
        round(stats::quantile(data2[[performance.test]],
          probs = 0.25
        ), 2),
        round(stats::median(data2[[performance.test]]), 2),
        round(mean(data2[[performance.test]]), 2),
        round(stats::sd(data2[[performance.test]]), 2),
        round(stats::quantile(data2[[performance.test]],
          probs = 0.75
        ), 2),
        round(max(data2[[performance.test]]), 2)
      ))
      tab.perftest <- cbind(tab.perftest, tab.perftest2)
      colnames(tab.perftest)[j] <- as.character(tab.freqvotat$Var1[j])
    }
  }

  tot.general <- as.matrix.data.frame(rbind(
    round(length(data[[performance.test]]), 0),
    round(min(data[[performance.test]]), 2),
    round(stats::quantile(data[[performance.test]],
      probs = 0.25
    ), 2),
    round(stats::median(data[[performance.test]]), 2),
    round(mean(data[[performance.test]]), 2),
    round(stats::sd(data[[performance.test]]), 2),
    round(stats::quantile(data[[performance.test]],
      probs = 0.75
    ), 2),
    round(max(data[[performance.test]]), 2)
  ))

  tab.perftest <- cbind(
    c(
      "Total N", "Min", "1st.Qu", "Median",
      "Mean", "SD", "3st.Qu", "Max"
    ),
    tot.general, tab.perftest
  )
  colnames(tab.perftest) <- c(
    "Statistics", "Total",
    colnames(tab.perftest)[3:length(colnames(tab.perftest))]
  )

  message(
    "\n Summary of ", performance.test, " by ", strategy.var,
    " - Individual level"
  )
  pander::pandoc.table(tab.perftest, split.tables = 100)
  polyserial.cor <- psych::polyserial(
    x = as.matrix(data[, performance.test]),
    y = as.matrix(data[, strategy.var])
  )
  message(paste0(
    "Biserial/Polyserial correlation = ",
    round(polyserial.cor[1], 4), "\n"
  ))
}