R/stats.R

In kahaaga/tstools: Useful functions

#' Computes a set of basic statistics for a scalar valued vector
#'
#' @param v A scalar valued vector.
#' @param percentiles Which percentiles to calculate for?
#' @param as.df Return as a data frame? Gives a named vector otherwise.
#' @return A named vector of statistical parameters of the vector 'v'
#'
#' @export
summary_stats <- function(v,
                          percentiles = c(0.01, 0.05, 0.1, 0.3,
                                          0.7, 0.9, 0.95, 0.99),
                          as.df = F) {
    if (class(v) == "data.frame") v <- as.vector(v)

    v.samplesize.without.nans <- length(v[stats::complete.cases(v)])
    v.samplesize <- length(v)
    v.var <- stats::var(v, na.rm = T)
    v.stdev <- stats::sd(v, na.rm = T)
    v.min <- min(v, na.rm = T)
    v.max <- max(v, na.rm = T)
    v.mean <- mean(v, na.rm = T)
    v.median <- stats::median(v, na.rm = T)
    v.skewness <- moments::skewness(v, na.rm = T)
    v.kurtosis <- moments::kurtosis(v, na.rm = T)
    v.mad <- stats::mad(v, na.rm = T)
    v.IQR <- stats::IQR(v, na.rm = T)
    Q1 <- stats::quantile(v, 0.25, na.rm = T)[[1]]
    Q3 <- stats::quantile(v, 0.75, na.rm = T)[[1]]
    boxplotmin <- Q1 - (1.5 * v.IQR)
    boxplotmax <- Q3 + (1.5 * v.IQR)
    boxplotnotchlower <- v.median - (1.58 * v.IQR /
                                       sqrt(v.samplesize.without.nans))
    boxplotnotchupper <- v.median + (1.58 * v.IQR /
                                       sqrt(v.samplesize.without.nans))

    # Compute quantiles
    percentiles <- sapply(X = percentiles,  FUN = function(p) {
                            stats::quantile(v, p, na.rm = T)
                        })

    v.outliers.fraction <- length(grDevices::boxplot.stats(v)) / v.samplesize

    summary.stats <- c("min" = v.min,
                      "max" = v.max,
                      "mean" = v.mean,
                      "median" = v.median,
                      "var" = v.var,
                      "stdev" = v.stdev,
                      "MAD" = v.mad,
                      "skewness" = v.skewness,
                      "kurtosis" = v.kurtosis,
                      "IQR" = v.IQR,
                      "Q1"  = Q1,
                      "Q3"  = Q3,
                      "boxplotmin" = boxplotmin,
                      "boxplotmax" = boxplotmax,
                      "boxplotnotchlower" = boxplotnotchlower,
                      "boxplotnotchupper" = boxplotnotchupper,
                      "samplesize" = v.samplesize,
                      "samplesize.without.nans" = v.samplesize.without.nans,
                      "outlier.fraction" = v.outliers.fraction)

    comb <- c(summary.stats, percentiles)
    if (as.df) return(as.data.frame(as.list(comb)))
    else return(comb)
}