TTR Yang-Zhang volatility bug?
What version of TTR are you using? I may have already fixed this on R-Forge. -- Joshua Ulrich | about.me/joshuaulrich FOSS Trading | www.fosstrading.com R/Finance 2013: Applied Finance with R | www.RinFinance.com
On Fri, Mar 15, 2013 at 2:49 PM, rex <rex at nosyntax.net> wrote:
I'm generating lognormal OHLC series and testing the RNG using TTR's
volatility functions. The yang.zhang calculation prepends too many NAs.
When Joshua's Yang-Zhang code is added as a function in my code, it
prepends fewer NAs, but still one too many, I think.
Why does what appears to be the same code behave differently?
R version 2.15.3 (2013-03-01) -- "Security Blanket"
Copyright (C) 2013 The R Foundation for Statistical Computing
ISBN 3-900051-07-0
Platform: x86_64-pc-linux-gnu (64-bit)
library(TTR)
library(xts)
vYZ = function(OHLC, n=10, N=252, ...){ #Joshua's Yang-Zhang TTR
function
# Historical Open-High-Low-Close Volatility: Yang Zhang
# http://www.sitmo.com/eq/417 #dead link
if(is.xts(OHLC)) {
Cl1 <- lag(OHLC[,4])
} else {
Cl1 <- c( NA, OHLC[-NROW(OHLC),4] )
}
dots <- list(...)
if(is.null(dots$alpha)) {
alpha <- 1.34
}
if(is.null(dots$k)) {
k <- (alpha-1) / ( alpha + (n+1)/(n-1) )
}
s2o <- N * runVar(log(OHLC[,1] / Cl1), n=n)
s2c <- N * runVar(log(OHLC[,4] / OHLC[,1]), n=n)
#s2rs <- volatility(OHLC=OHLC, n=n, calc="rogers.satchell", N=N, ...)
s2rs <- sqrt( N/n * runSum( #taken from Joshua's rogers.satchell
code
log(OHLC[,2]/OHLC[,4]) * log(OHLC[,2]/OHLC[,1]) +
log(OHLC[,3]/OHLC[,4]) * log(OHLC[,3]/OHLC[,1]), n ) )
s <- sqrt(s2o + k*s2c + (1-k)*(s2rs^2))
}
#my code starts here
options(width=132, digits=4)
ohlc = as.zoo(matrix(nrow=nDays,ncol=4))
colnames(ohlc) = c('Open','High','Low','Close')
allDays = seq(as.Date('2012-01-01'), as.Date('2012-02-24'), by = 1)
wkDays = allDays[!weekdays(allDays) %in% c('Saturday','Sunday')]
index(ohlc) = wkDays
ohlc = as.xts(ohlc)
tDays = 252 #trading days/year
nDays = length(wkDays) #random walk this many
trading days
hist = 15 #days used to estimate
volatility
drift = 0.10 #annual trend of walk
vol = 0.25 #25% annual volatility
stkPr = 100 #initial stock price
pricesDay = 6.5*12 #6.5 hours open and 12 prices
per hour
pricesWlk = pricesDay*nDays #number of prices in each
random walk
pricesYr = pricesDay*tDays #number of prices in a year
nWalks = 100 #number of random walks
tRands = pricesWlk*nWalks #number of lognormal rands
required
mu = drift/(pricesYr) #trend per trade for walk
sd = vol/sqrt(pricesYr) #volatility per trade for
walk expPr = stkPr*exp((drift +
0.5*vol*vol)*nDays/tDays) #expected end price
y = matrix(rlnorm(tRands, mu, sd), nWalks, pricesWlk) #random
lognormals
z = stkPr*t(apply(y,1,cumprod)) #do the day-by-day
multiplication
endPr = mean(z[,pricesWlk]) #average price at end of walk
sumYZ = sumCC = 0
for (w in 1:nWalks){
for (i in 1:nDays){
j = (i-1)*pricesDay + 1
k = j + pricesDay -1
ohlc[i,1] = z[w,j] #open for day
ohlc[i,2] = max(z[w,j:k]) #high for day
ohlc[i,3] = min(z[w,j:k]) #low for day
ohlc[i,4] = z[w,i*pricesDay] #close for day
}
volYZ2 = vYZ(ohlc, n=hist, N=tDays)
volYZ = volatility(ohlc, n=hist, N=tDays, calc="yang.zhang") #Yang-Zhang
volatility
volCC = volatility(ohlc, n=hist, N=tDays, calc="close") #closing
volatility
sumYZ = sumYZ + mean(na.omit(volYZ))
sumCC = sumCC + mean(na.omit(volCC))
#print(c(mean(volYZ),mean(volCC))) #there are multiple estimates
}
print(c('averageYZ', sumYZ/nWalks))
#[1] "averageYZ" "0.227380100996657"
print(c('averageCC', sumCC/nWalks))
#[1] "averageCC" "0.239352679348972"
#chartSeries(ohlc, name='SIM', theme='white')
volYZ #too many NAs
[,1]
2012-01-02 NA
2012-01-03 NA
2012-01-04 NA
2012-01-05 NA
2012-01-06 NA
2012-01-09 NA
2012-01-10 NA
2012-01-11 NA
2012-01-12 NA
2012-01-13 NA
2012-01-16 NA
2012-01-17 NA
2012-01-18 NA
2012-01-19 NA
2012-01-20 NA
2012-01-23 NA
2012-01-24 NA
2012-01-25 NA
2012-01-26 NA
2012-01-27 NA
2012-01-30 NA
2012-01-31 NA
2012-02-01 NA
2012-02-02 NA
2012-02-03 NA
2012-02-06 NA
2012-02-07 NA
2012-02-08 NA
2012-02-09 NA
2012-02-10 0.2125
2012-02-13 0.2042
2012-02-14 0.2010
2012-02-15 0.2011
2012-02-16 0.2077
2012-02-17 0.2133
2012-02-20 0.2222
2012-02-21 0.2261
2012-02-22 0.2270
2012-02-23 0.2233
2012-02-24 0.2271
volYZ2 #better
[,1]
2012-01-02 NA
2012-01-03 NA
2012-01-04 NA
2012-01-05 NA
2012-01-06 NA
2012-01-09 NA
2012-01-10 NA
2012-01-11 NA
2012-01-12 NA
2012-01-13 NA
2012-01-16 NA
2012-01-17 NA
2012-01-18 NA
2012-01-19 NA
2012-01-20 NA
2012-01-23 0.2067
2012-01-24 0.2087
2012-01-25 0.2100
2012-01-26 0.2110
2012-01-27 0.2139
2012-01-30 0.2138
2012-01-31 0.2091
2012-02-01 0.2163
2012-02-02 0.2240
2012-02-03 0.2272
2012-02-06 0.2245
2012-02-07 0.2259
2012-02-08 0.2229
2012-02-09 0.2183
2012-02-10 0.2151
2012-02-13 0.2069
2012-02-14 0.2038
2012-02-15 0.2038
2012-02-16 0.2099
2012-02-17 0.2155
2012-02-20 0.2229
2012-02-21 0.2266
2012-02-22 0.2292
2012-02-23 0.2250
2012-02-24 0.2285
volCC
[,1]
2012-01-02 NA
2012-01-03 NA
2012-01-04 NA
2012-01-05 NA
2012-01-06 NA
2012-01-09 NA
2012-01-10 NA
2012-01-11 NA
2012-01-12 NA
2012-01-13 NA
2012-01-16 NA
2012-01-17 NA
2012-01-18 NA
2012-01-19 NA
2012-01-20 0.2288
2012-01-23 0.2281
2012-01-24 0.2628
2012-01-25 0.2618
2012-01-26 0.2715
2012-01-27 0.2710
2012-01-30 0.2766
2012-01-31 0.2603
2012-02-01 0.2891
2012-02-02 0.3014
2012-02-03 0.3083
2012-02-06 0.2822
2012-02-07 0.2812
2012-02-08 0.2813
2012-02-09 0.2788
2012-02-10 0.2845
2012-02-13 0.2471
2012-02-14 0.2654
2012-02-15 0.2560
2012-02-16 0.2827
2012-02-17 0.2627
2012-02-20 0.2832
2012-02-21 0.2686
2012-02-22 0.3048
2012-02-23 0.3093
2012-02-24 0.3536
Thanks,
-rex
--
_______________________________________________ R-SIG-Finance at r-project.org 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.