rquantitative-financeperformanceanalytics
PerformanceAnalytics R package's historic VaR doesn't match my calcs
I computed historic VaR on a simple equity portfolio (for simplicity's sake, let's say there are only two stocks with one year of daily history). Then I did it again using the PerformanceAnalytics R package. My two results don't match.
Am I missing something on how to do basic VaR, or is PerformanceAnalytics doing something strange?
My sample code is below:
library(xts)
# (I'll define "sr" at the end of the question. It's a big xts structure)
weights = c(0.7929511, 0.2070489)
p = 0.95
# How I do it:
m = data.matrix(as.data.frame(sr))
daily_whatif_scenario = tcrossprod(weights, m)[1,]
my_historic_VaR = as.numeric(-quantile(daily_whatif_scenario, 1 - p)[1])
# How PerformanceAnalytics does it:
their_historic_VaR = PerformanceAnalytics::VaR(R = sr, p = p,
method = "historical", portfolio_method = "component", weights = weights)
My calculation comes up with 1.325%, but the package gives me 1.395%. Why such a big difference? These differences only get bigger as I add more stocks and history to my example. Am I doing something wrong?
Here is the result of dput(sr):
sr = structure(c(-0.0162, 0.0049, 0.0029, -1e-04, -0.0079, -0.0101,
0.0053, 0.007, -0.0107, 0.001, -0.003, -0.0033, 0.0013, 0.0163,
0.0102, 2e-04, -0.0027, 0.0303, -0.0149, 0.0034, 0.0073, 0.0111,
-0.0057, 0.0022, 0.0266, 0.0038, 0.0062, -0.0063, 0.0078, 0.0022,
0.0091, -0.0074, 0.008, 0.0068, 0.0048, 0.0031, 0.0142, -0.0034,
-0.0089, 0.001, 0.001, -0.0141, 0.0042, 0.0026, -0.017, 0.001,
-0.006, 0.0048, 0.0064, -0.0063, -0.0122, 0.0122, -0.0126, 0.0239,
-0.0236, -9e-04, 0.004, -0.0049, 0.0028, -0.0109, 0.0106, 0.0066,
0.0096, 0.0075, 0.0082, -0.008, 0.047, -6e-04, 5e-04, -0.005,
-0.0084, -0.0074, 0.0078, -0.0029, 0.0071, 5e-04, 0.0118, 0.0068,
-0.0032, 0.0068, -0.0011, -0.0095, 0.0029, -0.0019, -0.0084,
0.0065, -0.014, 0.0049, -0.0042, 0.0071, -0.0097, -0.0038, -0.0049,
0.0045, 0.001, -7e-04, -0.0056, 0.0098, -0.0112, 0.0093, 5e-04,
-6e-04, 0, -0.0013, 0.0026, 0.0199, -8e-04, 0.0023, -0.0038,
-0.0035, -0.0045, -0.0323, -0.0021, -0.0054, -0.0102, 0.0222,
-0.0139, -0.0027, 6e-04, -5e-04, -0.0069, -8e-04, 0.0074, -0.0071,
0.0017, -0.0142, 0.0111, 0.0022, -0.0085, 0.0183, -0.0061, -0.0031,
0.0083, 0.0111, -0.0028, -0.0051, -0.0149, 0.0026, -0.0046, -0.0105,
-0.003, 0.0019, 0.0055, 0.0471, 0.0042, 9e-04, -0.0226, -0.0175,
0.0169, 3e-04, 8e-04, -0.0258, -0.0047, -0.0044, 0.0071, 0.0129,
0.0063, 0.0248, 0.0107, 0.0133, -0.0214, 0.0045, 0.0175, -0.0034,
0.0048, -1e-04, -0.0118, 0.0082, -0.0024, 0.0019, -0.0139, 0.0037,
-0.0018, -0.0166, -0.002, -0.0058, 0.0053, -0.0086, 0.0159, 0.0027,
-0.0097, 0.0186, -0.0099, 0.0362, 0.0034, -0.0044, 0.005, -0.0074,
0.0031, 0.0151, 0, -0.0036, -0.0054, -0.004, -0.0088, 0, 0.015,
-0.0019, 3e-04, 0.0087, -0.0017, 0.0015, 0.0046, -0.0177, 0.0018,
0.0021, -0.0142, 4e-04, -0.005, -0.001, -0.0108, -5e-04, 0.0167,
0, -7e-04, -0.0079, 0.0397, -6e-04, -0.0132, -0.0013, -0.0022,
0.001, -0.0021, 0.0119, -0.0031, 0.0065, 0.001, -0.0014, -0.0054,
8e-04, -0.0021, 0.0103, -0.0016, -0.0014, 0.0169, -4e-04, -0.0067,
0.0249, 0.0018, 0.0047, -0.0117, -0.0015, 0.0133, 0.0144, -3e-04,
-0.0084, -0.0067, 0.0048, -0.0226, 0.001, -0.0079, -1e-04, 0.0062,
0.0037, 0, 7e-04, -0.0046, -0.0045, -7e-04, 0.0097, -0.0066,
-0.0054, -0.0029, -0.0041, -0.0113, 0.0059, -0.0134, -0.0177,
0.0057, -0.0043, -0.0106, -0.0115, -0.0198, 0.0131, 0.0086, -0.008,
0.0286, -0.0042, 6e-04, 0.0151, 0.0071, 0.0058, 0.0056, 0.0165,
0.0143, 0.0092, 0.0023, 0.0168, 0.0068, -0.0033, -0.0296, -5e-04,
0.0191, -0.0106, 0.0062, -0.002, -0.001, 0.0038, 0.0039, 0.001,
-0.01, 0.0059, 0.0036, -0.0063, -0.0082, -0.0034, 0.0039, 0.0184,
0.0055, -0.0084, 0.0093, -0.0104, 0.0289, -0.01, -0.0098, 0.0108,
-0.0068, -0.0056, 0.002, 0.0206, -0.008, 0.0212, 0.0496, 0.0038,
-0.008, 0.0122, 0.0029, -0.0014, -9e-04, -0.0077, -0.016, 0.0064,
0.0012, 0.0056, 0.0039, 0.0176, -0.0125, -0.0088, 0.0075, 0.0049,
-0.0131, -0.004, 0.0052, -0.004, 0.0021, -0.0162, -0.0126, 0.0042,
0.0032, -0.0059, 0.0014, -0.0034, 7e-04, 0.0043, -6e-04, -0.0067,
0.0067, 0.0045, 0.0096, -3e-04, -0.0082, 0.0038, 0.0053, -0.013,
0.0197, -0.0057, 0.0041, -0.0142, 0.0114, -0.0045, -0.0123, -0.0433,
-0.0066, -0.0212, 0.0257, -0.0068, 8e-04, -0.0031, -0.0029, -0.0097,
-0.0074, 0.0065, -0.013, -0.0036, -0.013, -0.002, 0.0032, -0.0087,
0.0054, -0.001, -0.0118, 0.0034, 0.0051, -0.0065, 2e-04, 0.0015,
-3e-04, 0.0043, -0.0168, 8e-04, -0.0033, -0.0094, -0.0017, -4e-04,
0.0024, -3e-04, -0.0087, 0.0073, 0.0176, 0.0014, -0.0121, -0.005,
-0.006, -0.0059, 0.0159, 0.0273, -0.0071, -0.0163, 0.0041, 0.004,
-6e-04, 0.0095, 0.0012, 0.0156, 0.003, 0.0153, -0.0203, -0.0024,
0.0101, -0.0108, 0.0118, -0.0172, -0.0077, -2e-04, 0, 2e-04,
0.0287, -0.0023, 0.0096, 0.0043, -0.0016, -0.0169, 0.0221, 0.0208,
-0.0036, -0.0236, 0.014, 0.0047, 0.0023, 0, -0.0013, -0.0033,
0.0024, -0.0134, 0, 0.0081, -0.0085, -0.006, -0.0014, -0.0184,
-0.0066, 0.0103, -0.0123, -0.0022, 0.0021, 0.0059, -0.0014, -0.0046,
0.0124, -0.009, 0.0079, 0.0096, -0.0019, -0.0025, -0.0024, -0.0063,
-0.0139, -0.0094, 0.0152, -0.0145, 0.0061, 0.0055, 0.016, -0.0105,
0.0061, 0.0053, -0.0058, -0.0272, -0.0345, -0.0021, 0.0322, -0.0092,
-0.0022, 0.0154, -0.0243, 0.0072, 0.0078, 0.0152, -7e-04, -0.0081,
-0.0042, 0.0038, 0.0157, -0.0064, -0.0063, -3e-04, 0.0094, -0.0162,
-0.0045, -0.0122, -0.0062, -0.0081, 0.0112, -1e-04, -0.0059,
9e-04, 0.0043, -0.0081), .indexCLASS = "Date", tclass = c("POSIXct",
"POSIXt"), .indexTZ = "", tzone = "", class = c("xts", "zoo"), index = structure(c(1459728000,
1459814400, 1459900800, 1459987200, 1460073600, 1460332800, 1460419200,
1460505600, 1460592000, 1460678400, 1460937600, 1461024000, 1461110400,
1461196800, 1461283200, 1461542400, 1461628800, 1461715200, 1461801600,
1461888000, 1462147200, 1462233600, 1462320000, 1462406400, 1462492800,
1462752000, 1462838400, 1462924800, 1463011200, 1463097600, 1463356800,
1463443200, 1463529600, 1463616000, 1463702400, 1463961600, 1464048000,
1464134400, 1464220800, 1464307200, 1464566400, 1464652800, 1464739200,
1464825600, 1464912000, 1465171200, 1465257600, 1465344000, 1465430400,
1465516800, 1465776000, 1465862400, 1465948800, 1466035200, 1466121600,
1466380800, 1466467200, 1466553600, 1466640000, 1466726400, 1466985600,
1467072000, 1467158400, 1467244800, 1467331200, 1467590400, 1467676800,
1467763200, 1467849600, 1467936000, 1468195200, 1468281600, 1468368000,
1468454400, 1468540800, 1468800000, 1468886400, 1468972800, 1469059200,
1469145600, 1469404800, 1469491200, 1469577600, 1469664000, 1469750400,
1470009600, 1470096000, 1470182400, 1470268800, 1470355200, 1470614400,
1470700800, 1470787200, 1470873600, 1470960000, 1471219200, 1471305600,
1471392000, 1471478400, 1471564800, 1471824000, 1471910400, 1471996800,
1472083200, 1472169600, 1472428800, 1472515200, 1472601600, 1472688000,
1472774400, 1473033600, 1473120000, 1473206400, 1473292800, 1473379200,
1473638400, 1473724800, 1473811200, 1473897600, 1473984000, 1474243200,
1474329600, 1474416000, 1474502400, 1474588800, 1474848000, 1474934400,
1475020800, 1475107200, 1475193600, 1475452800, 1475539200, 1475625600,
1475712000, 1475798400, 1476057600, 1476144000, 1476230400, 1476316800,
1476403200, 1476662400, 1476748800, 1476835200, 1476921600, 1477008000,
1477267200, 1477353600, 1477440000, 1477526400, 1477612800, 1477872000,
1477958400, 1478044800, 1478131200, 1478217600, 1478476800, 1478563200,
1478649600, 1478736000, 1478822400, 1479081600, 1479168000, 1479254400,
1479340800, 1479427200, 1479686400, 1479772800, 1479859200, 1479945600,
1480032000, 1480291200, 1480377600, 1480464000, 1480550400, 1480636800,
1480896000, 1480982400, 1481068800, 1481155200, 1481241600, 1481500800,
1481587200, 1481673600, 1481760000, 1481846400, 1482105600, 1482192000,
1482278400, 1482364800, 1482451200, 1482710400, 1482796800, 1482883200,
1482969600, 1483056000, 1483315200, 1483401600, 1483488000, 1483574400,
1483660800, 1483920000, 1484006400, 1484092800, 1484179200, 1484265600,
1484524800, 1484611200, 1484697600, 1484784000, 1484870400, 1485129600,
1485216000, 1485302400, 1485388800, 1485475200, 1485734400, 1485820800,
1485907200, 1485993600, 1486080000, 1486339200, 1486425600, 1486512000,
1486598400, 1486684800, 1486944000, 1487030400, 1487116800, 1487203200,
1487289600, 1487548800, 1487635200, 1487721600, 1487808000, 1487894400,
1488153600, 1488240000, 1488326400, 1488412800, 1488499200, 1488758400,
1488844800, 1488931200, 1489017600, 1489104000, 1489363200, 1489449600,
1489536000, 1489622400, 1489708800, 1489968000, 1490054400, 1490140800,
1490227200, 1490313600, 1490572800, 1490659200, 1490745600, 1490832000
), tclass = "Date"), .Dim = c(259L, 2L), .Dimnames = list(NULL,
c("DHR.US", "GIS.US")))
Solution
So I dug up the code that is run when VaR is called with the parameters you specified. Please see below.
As you read my answer please note that this domain is out side of my skill set and the math is a bit over my head, however it appears that the major difference between your code and theirs is they sort the data and select the transposed cross product identified by alpha. Where as you use quantile. For some reason this means you select adjacent numbers - it appears to be basically a rounding error. Please see the following excel screen shot of your daily_whatif_scenario to illustrate this point:
Code for VaR:
VaR.historical.portfolio = function(R,p,w)
{
alpha = .setalphaprob(p)
portret = c();
T = dim(R)[1]
N = dim(R)[2]
for( t in c(1:T) ){
portret = c(portret,sum(w*as.numeric(R[t,])))
}
hVaR = -1* sort(portret)[floor(alpha*T)]
return(hVaR)
}
Where:
.setalphaprob = function (p)
{
if ( p >= 0.51 ) {
# looks like p was a percent like .99
alpha <- 1-p
} else {
alpha = p
}
return (alpha)
}
- Determining as to whether the function call is stored in a variable
- How to increase resolution of ggplots without using ggsave in officer?
- How to apply a function within a for loop for a set of existing variables of a predifined structure in R?
- How to change the number of decimal places of the mean in a data frame
- Forestplot in R superscript in tabletext matrix
- Converting colnames from character to numeric when extracting them from a dataframe
- With different seeds can Random States repeat in R
- How to Stack Painted Phylogenetic Trees in R Like ggdensitree but with Colored Regimes
- How to import/rbind multiple files with different indexed columns?
- Pivot wider in R return list instead of data frame
- R workspace doesn't load anything in Visual Studio Code
- Search for multiple occurrences of substring within string
- Looking for faster way to implement logSumExp across multidimensional array
- How to get \bm{} to work in an R markdown (to HTML) file?
- Leaving a space in geom_bar with stat='bin', position='dodge' and zeroes
- Where are the 1st stage F statistics located in a "fixest" object?
- Collapsing data per family
- How to set default template for new ".R" files in rstudio
- Adding two-way fixed effects for zero inflated Poisson model
- How to Move R Code and Output to the Right in a Quarto PDF Document?
- What does runif() mean when used inside if_else()?
- Remove columns from dataframe where ALL values are NA
- How to pass unquoted argument to filter() within user defined function
- Optimizing the distribution of random integers across data frame rows
- R dplyr conditional join with "join_by" not working
- How to hide a reappearing hover label for a trace in R-Plotly within Shiny?
- Estimating non-monotonic bi-exponential curve fit
- Problem minimising a function using L-BFGS-B method in R?
- trying to obtain TRUE/FALSE response to the question "is argument name x valid for function y?"
- Retrieve whole lyrics from URL