Rolling Regression Data Frame - r

Question

Appreciate this may have been asked before but I have not found a clear solution to work over a data frame.
I want to run a rolling linear regression over a look back of 5 days. (small so can illustrate here)
So far I am trying:
rollingbeta <- rollapply(df,
width=5,
FUN = function(Z)
{
t = lm(formula=y_Close ~ x_Close+0, data = as.data.frame(Z));
return(t$coef)[1]
},
by.column=FALSE, align="right",fill = NA)
head(rollingbeta,100)
However, I expect to have the beta for the rolling lookback window. Instead I have and output with 10 columns.
> NCOL(rollingbeta)
[1] 10
Can anyone assist?
Here is dummy data (save to .txt and read)
df <- read.table("your_dir\df.txt",header=TRUE, sep="", stringsAsFactors=FALSE)
Date open.x high.x low.x x_Close volume.x open.y high.y low.y y_Close volume.y x.y.cor
1451 2010-01-04 57.32 58.13 57.32 57.85 442900 6.61 6.8400 6.61 6.83 833100 NA
1452 2010-01-05 57.90 58.33 57.54 58.20 436900 6.82 7.1200 6.80 7.12 904500 NA
1453 2010-01-06 58.20 58.56 58.01 58.42 850600 7.05 7.3800 7.05 7.27 759800 NA
1454 2010-01-07 58.31 58.41 57.14 57.90 463600 7.24 7.3000 7.06 7.11 557800 NA
1455 2010-01-08 57.45 58.62 57.45 58.47 206500 7.08 7.3500 6.95 7.29 588100 NA
1456 2010-01-11 58.79 59.00 57.22 57.73 331900 7.38 7.4500 7.17 7.22 450500 NA
1457 2010-01-12 57.20 57.21 56.15 56.34 428500 7.15 7.1900 6.87 7.00 694700 NA
1458 2010-01-13 56.32 56.66 54.83 56.56 577500 7.05 7.1700 6.98 7.15 528800 NA
1459 2010-01-14 56.51 57.05 55.37 55.53 368100 7.08 7.1701 7.08 7.11 279900 NA
1460 2010-01-15 56.59 56.59 55.19 55.84 417900 7.03 7.0500 6.95 7.03 407600 NA
The output should for the first rolling linear regression should be:
NA NA NA NA NA 0.1229065

Answer 1

A faster alternative than wibeasley's answer is to use the rollRegres package as follows
ds <- structure(list(Date = structure(
c(14613, 14614, 14615, 14616, 14617, 14620, 14621, 14622, 14623, 14624), class = "Date"),
open.x = c(57.32, 57.9, 58.2, 58.31, 57.45, 58.79, 57.2, 56.32, 56.51, 56.59),
high.x = c(58.13, 58.33, 58.56, 58.41, 58.62, 59, 57.21, 56.66, 57.05, 56.59),
low.x = c(57.32, 57.54, 58.01, 57.14, 57.45, 57.22, 56.15, 54.83, 55.37, 55.19),
x_Close = c(57.85, 58.2, 58.42, 57.9, 58.47, 57.73, 56.34, 56.56, 55.53, 55.84),
volume.x = c(442900L, 436900L, 850600L, 463600L, 206500L, 331900L, 428500L, 577500L, 368100L, 417900L),
open.y = c(6.61, 6.82, 7.05, 7.24, 7.08, 7.38, 7.15, 7.05, 7.08, 7.03),
high.y = c(6.84, 7.12, 7.38, 7.3, 7.35, 7.45, 7.19, 7.17, 7.1701, 7.05),
low.y = c(6.61, 6.8, 7.05, 7.06, 6.95, 7.17, 6.87, 6.98, 7.08, 6.95),
y_Close = c(6.83, 7.12, 7.27, 7.11, 7.29, 7.22, 7, 7.15, 7.11, 7.03),
volume.y = c(833100L, 904500L, 759800L, 557800L, 588100L, 450500L, 694700L, 528800L, 279900L, 407600L)),
row.names = c(NA, -10L), class = "data.frame")
# we get the same
library(roll)
library(rollRegres)
X <- as.matrix(ds$x_Close)
Y <- ds$y_Close
Ymat <- as.matrix(Y)
all.equal(
roll_lm(x = X, y = Ymat, intercept = FALSE, width = 5L)$coefficients[, 2],
drop(roll_regres.fit(x = X, y = Y, width = 5L)$coefs),
check.attributes = FALSE)
#R [1] TRUE
You can also fit the model with a formula as with lm using the roll_regres function
all.equal(
roll_lm(x = X, y = Ymat, intercept = FALSE, width = 5L)$coefficients[, 2],
drop(roll_regres(y_Close ~ x_Close - 1, ds, width = 5L)$coefs),
check.attributes = FALSE)
#R [1] TRUE
Here is a benchmark of the computation speed
# We add a few more observation to get an interesting example
set.seed(1)
n <- 250 * 5 # 5 years of trading data
X <- as.matrix(rnorm(n))
Y <- rnorm(n)
Ymat <- as.matrix(Y)
microbenchmark::microbenchmark(
roll_lm(x = X, y = Ymat, intercept = FALSE, width = 5L),
roll_regres.fit(x = X, y = Y, width = 5L),
times = 1e3)
#R Unit: microseconds
#R expr min lq mean median uq max neval
#R roll_lm(x = X, y = Ymat, intercept = FALSE, width = 5L) 663.7 739.9 834.2 777.1 860.2 3972.3 1000
#R roll_regres.fit(x = X, y = Y, width = 5L) 186.9 204.6 237.4 224.8 248.3 546.4 1000

Answer 2

Consider using the roll package.
library(magrittr); requireNamespace("roll")
ds <- readr::read_csv(
" Date, open.x, high.x, low.x, x_Close, volume.x, open.y, high.y, low.y, y_Close, volume.y
2010-01-04, 57.32, 58.13, 57.32, 57.85, 442900, 6.61, 6.8400, 6.61, 6.83, 833100
2010-01-05, 57.90, 58.33, 57.54, 58.20, 436900, 6.82, 7.1200, 6.80, 7.12, 904500
2010-01-06, 58.20, 58.56, 58.01, 58.42, 850600, 7.05, 7.3800, 7.05, 7.27, 759800
2010-01-07, 58.31, 58.41, 57.14, 57.90, 463600, 7.24, 7.3000, 7.06, 7.11, 557800
2010-01-08, 57.45, 58.62, 57.45, 58.47, 206500, 7.08, 7.3500, 6.95, 7.29, 588100
2010-01-11, 58.79, 59.00, 57.22, 57.73, 331900, 7.38, 7.4500, 7.17, 7.22, 450500
2010-01-12, 57.20, 57.21, 56.15, 56.34, 428500, 7.15, 7.1900, 6.87, 7.00, 694700
2010-01-13, 56.32, 56.66, 54.83, 56.56, 577500, 7.05, 7.1700, 6.98, 7.15, 528800
2010-01-14, 56.51, 57.05, 55.37, 55.53, 368100, 7.08, 7.1701, 7.08, 7.11, 279900
2010-01-15, 56.59, 56.59, 55.19, 55.84, 417900, 7.03, 7.0500, 6.95, 7.03, 407600"
)
runs <- roll::roll_lm(
x = as.matrix(ds$x_Close),
y = as.matrix(ds$y_Close),
width = 5,
intercept = FALSE
)
# Nested in a named-column, within a matrix, within a list.
ds$beta <- runs$coefficients[, "x1"]
ds$beta
# [1] NA NA NA NA 0.1224813
# [6] 0.1238653 0.1242478 0.1246279 0.1256553 0.1259121
Double-check the alignment of the variables in your dataset. x_Close is around 50, while y_Close is around 7. That might explain the small disparity between the expected 0.1229065 and the 0.1224813 value above.