R/returns_diagnostics_plots.R

In pawel-wieczynski/PolishStock: Analysis of WSE stocks and other assets.

#' Returns Plot function
#'
#' Plot histogram of returns alongside with empirical density, fitted normal density and fitted t-Student density. Plot ACF and qq-plots.
#' @param returns Numeric vector of asset returns

plot_histogram_with_density <- function(returns){

  # Check if returns is numeric vector
  if (class(returns) != 'numeric'){
    stop('returns is not numeric vector')
  }

  params <- returns_fit(returns)

  n <- 10*length(returns)
  sample_density <- density(returns)
  t_density <- density(params$t_param[1] +
                         rt(n, params$t_param[3]) *
                         params$t_param[2])
  normal_density <- density(rnorm(n, params$norm_param[1]
                                  ,params$norm_param[2]))


  hist(returns
       ,breaks = 'FD'
       ,freq = FALSE
       ,xlab = ''
       ,ylab = '')
  lines(sample_density
        ,col = 'blue'
        ,lwd = 3)
  lines(t_density
        ,col = 'orange'
        ,lwd = 3)
  lines(normal_density
        ,col = 'forestgreen'
        ,lwd = 3)
  legend('left'
         ,legend = c('Sample density'
                     ,'t-Student density'
                     ,'Normal density')
         ,col = c('blue'
                  ,'orange'
                  ,'forestgreen')
         ,bg = 'transparent'
         ,lty = c(1,1,1)
         ,lwd = c(3,3,3)
         ,box.lty = 0)

}

plot_qq <- function(returns){

  # Check if returns is numeric vector
  if (class(returns) != 'numeric'){
    stop('returns is not numeric vector')
  }

  params <- returns_fit(returns)

  n <- length(returns)

  # t-student QQ-plot
  xx <- quantile(params$t_param[1] +
                   qt(seq(0, 1, by = 0.01)
                      ,df = params$t_param[3]) *
                   params$t_param[2]
                 ,c(0.25, 0.75))
  yy <- quantile(returns
                 ,c(0.25, 0.75))
  slope <- (yy[2] - yy[1]) / (xx[2] - xx[1])
  intercept <- yy[1] - slope*xx[1]

  qqplot(x = params$t_param[1] +
           qt(seq(0, 1, length.out = n)
              ,df = params$t_param[3]) *
           params$t_param[2]
         ,y = returns
         ,col = 'blue'
         ,xlab = ''
         ,ylab = ''
         ,main = 't-Student QQ-plot')
  abline(a = intercept
         ,b = slope
         ,col = 'orange'
         ,lwd = 3)

  # Normal QQ-plot
  xx <- quantile(qnorm(seq(0, 1, by = 0.01)
                       ,mean = params$norm_param[1]
                       ,sd = params$norm_param[2])
                 ,c(0.25, 0.75))
  yy <- quantile(returns
                 ,c(0.25, 0.75))
  slope <- (yy[2] - yy[1]) / (xx[2] - xx[1])
  intercept <- yy[1] - slope*xx[1]

  qqplot(x = qnorm(seq(0, 1, length.out = n)
                   ,mean = params$norm_param[1]
                   ,sd = params$norm_param[2])
         ,y = returns
         ,col = 'blue'
         ,xlab = ''
         ,ylab = ''
         ,main = 'Normal QQ-plot')
  abline(a = intercept
         ,b = slope
         ,col = 'orange'
         ,lwd = 3)

}

diagnostic_plots <- function(returns, lags = 10){

  # Check if returns is numeric vector
  if (class(returns) != 'numeric'){
    stop('returns is not numeric vector')
  }

  par(mfcol = c(2,2))

  plot_histogram_with_density(returns)

  acf(returns
      ,lag.max = lags
      ,main = 'Autocorrelation of returns')

  plot_qq(returns)

  par(mfcol = c(1,1))

}