question about subsetting xts, calculating returns

[Andreas Voellenklee]

Hello,

I would like to monitor short-term-returns on simple strategies like
donchian channel price outbreaks. That is, I would like to see the
returns at 1:10 periods after the breakout-event occurs.

For example:
getSymbols("NOK", adjust=TRUE)

# calculate donchian channel for NOK
DC <- DonchianChannel(lag(cbind(Hi(NOK),Lo(NOK))))

# mark days when price closes below the low band of DC
breakdown <- eval(Cl(NOK)<DC[,"low"])
colnames(breakdown) <- c("breakdown")

# how often did that condition occur?
colSums(breakdown, na.rm=TRUE)
#breakdown
#      135


Now I masked out all days when the price closed below the donchian channel.

- Is there a easy way to display only the (135) rows where
breakdown[,1] is TRUE?
- How to calculate the 1:10 days- returns after the breakdown-events then?

I would be grateful for some hints,

Andreas

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should go.

[gsee]

Hi Andreas,

Try something like below.  Make sure it's right though, because I
didn't really check my work.

On Thu, Apr 12, 2012 at 11:36 AM, Andreas Voellenklee
<[hidden email]> wrote:
> - Is there a easy way to display only the (135) rows where
> breakdown[,1] is TRUE?

NOK[breakdown]

> - How to calculate the 1:10 days- returns after the breakdown-events then?

# 10 period return from the current period through 10 periods into the future
lag(ROC(Cl(NOK), 10), -10)[breakdown]

# 1:10 period return from the current period through 1:10 periods into
the future
do.call(cbind, lapply(1:10, function(i) {
    lag(ROC(Cl(NOK), i), -i)[breakdown]
}))


HTH,
Garrett

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should go.

[Andreas Voellenklee]

Thank you Sean and Garrett. Garrett's example worked fine. If anybody
is interested, I incorporated this in a function and wrote some more
to visualize the results.


calculateNPeriodReturns <- function (TS, Filter, n)
{
    returns <- do.call(cbind, lapply(1:n, function(i) {
        lag(ROC(Cl(TS), i), -i)[Filter]
    }))
    colnames(returns) <- 1:n
    return(returns)
}

returnsBoxplot <- function (returns)
{
    par(mfrow = c(1, 1))
    boxplot(as.matrix(returns), range = 0)
}

returnsHistogram <- function (returns)
{
    columns <- dim(returns)[2]/2
    par(mfrow = c(2, columns))
    for (i in 1:dim(returns)[2]) {
        hist(returns[, i], main = paste("t+", i, sep = ""), xlab = "",
            breaks = 20, col = "blue")
    }
}

returnsLinePlot <- function (returns)
{
    molten <- melt(t(as.data.frame(returns)))
    colnames(molten) <- c("period", "date", "return")
    ggplot(molten, aes(period, return)) + geom_line(aes(colour = date))
}

library(quantmod)
# try Nokia for this example
getSymbols("NOK", adjust=TRUE)

# calculate donchian channel for NOK
DC <- DonchianChannel(lag(cbind(Hi(NOK),Lo(NOK))))

# mark days when price closes below the low band of DC
breakdown <- eval(Cl(NOK)<DC[,"low"])
colnames(breakdown) <- c("breakdown")

# calculate returns from 1:10 periods after breakdown events
returns <- calculateNPeriodReturns(NOK, breakdown, 10)

# show a boxplot of returns
returnsBoxplot(returns)

# show a histogram of returns
returnsHistogram(returns)

# show a line diagram
library(ggplot2)
library(reshape)
returnsLinePlot(returns)

_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
-- Subscriber-posting only. If you want to post, subscribe first.
-- Also note that this is not the r-help list where general R questions should go.