Again with variability of long-short decile tests

December 6, 2012

A simpler approach to producing the variability.

Previously

The post “Variability in long-short decile strategy tests” proposed a way of assessing the variability of strategy tests in which a long-short portfolio is created by equally weighting the top and bottom deciles.

Improved idea

Joe Mezrich suggests maintaining equal weights but bootstrapping the assets within the deciles.  For me that passes the jealousy test — why didn’t I think of that?

It eliminates the arbitrariness of the selection of the range of weights to use in the previous method.

Figure 1 shows the variability of 1000 bootstrap paths for the MACD signal that the previous post used.

Figure 1: Efficacy of MACD via long-short deciles with bootstrapped equal weights. The bootstrap variability is bigger than the biggest of the variabilities investigated previously.

Appendix R

The function to do the bootstrapping is a very mildly revised version of the function that does the random weighting.

pp.decileTestBoot <-
  function(signal, prices, trials=1000, groups=10)
  {
    # R function to test a signal 
    # via long-short deciles
    # and bootstrapping the deciles

    # put in the public domain 2012 by Burns Statistics

    # testing status:
    # seems to work

    stopifnot(all(dim(signal) == dim(prices)), 
              length(groups) == 1,
              identical(sort(colnames(signal)), 
              sort(colnames(prices))))

    ntimes <- nrow(prices)
    eqwtval <- rep(NA, ntimes)
    names(eqwtval) <- rownames(prices)
    randwtval <- array(NA, c(length(eqwtval), trials),
                       list(names(eqwtval), NULL))
    uret <- tail(prices, -1) / head(prices, -1) - 1
    randwtval[1,] <- eqwtval[1] <- 100

    nside <- round(ncol(prices) / groups)

    if(trials) {
      tseq <- 1:trials
      t.eret <- numeric(trials)
    }

    for(i in 1:(ntimes-1)) {
      tb <- pp.topBottom(signal[i, ], n=nside)
      botret <- uret[i, tb$bottom]
      topret <- uret[i, tb$top]
      this.eret <- mean(topret) - mean(botret)
      eqwtval[i+1] <- eqwtval[i] * (1 + this.eret)
      if(trials) {
        for(j in tseq) {
          t.eret[j] <- mean(sample(topret, nside, 
            replace=TRUE)) - mean(sample(botret, 
            nside, replace=TRUE))
        }
        randwtval[i+1, ] <- randwtval[i, ] * 
            (1 + t.eret)
      }
    }
    ans <- list(equal.weight=eqwtval, 
        random.weight=randwtval, nside=nside, 
        call=match.call())
    class(ans) <- "SignalTest"
    ans
  }

Leave a Reply

  1. […] ← Popular posts 2012 November Again with variability of long-short decile tests → […]

Related posts

  • April 18, 2011

    How well do asset weight constraints constrain risk? The setup In "Unproxying weight constraints" I claimed that many constraints on asset weights are really a proxy for constraining risk. [...]

  • March 28, 2011

    Overfitting is a problem when trying to predict financial returns.  Perhaps you've heard that before.  Some simple examples should clarify what overfitting is -- and may surprise you. Polynomials [...]

  • March 7, 2011

    In "What the hell is a variance matrix?" I talked about the basics of variance matrices and highlighted challenges for estimating them in finance.  Here we look more deeply [...]