我有一个很大的'data.table'，由多个股票的每日回报组成。
rm(list = ls())
library(data.table)
set.seed(1080)
Firm1 <- rbind(data.table(Month = rep(200001, 3), Firm = rep(1, 3), Rt = rnorm(3)) ,data.table(Month = rep(200002, 4), Firm = rep(1, 4), Rt = rnorm(4)),
data.table(Month = rep(200003, 5), Firm = rep(1, 5), Rt = rnorm(5)), data.table(Month = rep(200004, 2), Firm = rep(1, 2), Rt = rnorm(2)),
data.table(Month = rep(200005, 3), Firm = rep(1, 3), Rt = rnorm(3)))
Firm2 <- rbind(data.table(Month = rep(200001, 1), Firm = rep(2, 1), Rt = rnorm(1)) , data.table(Month = rep(200002, 3), Firm = rep(2, 3), Rt = rnorm(3)),
data.table(Month = rep(200003, 4), Firm = rep(2, 4), Rt = rnorm(4)), data.table(Month = rep(200004, 1), Firm = rep(2, 1), Rt = rnorm(1)),
data.table(Month = rep(200005, 3), Firm = rep(2, 3), Rt = rnorm(3)))
Firm3 <- rbind(data.table(Month = rep(200001, 3), Firm = rep(3, 3), Rt = rnorm(3)) ,data.table(Month = rep(200002, 6), Firm = rep(3, 6), Rt = rnorm(6)),
data.table(Month = rep(200003, 5), Firm = rep(3, 5), Rt = rnorm(5)), data.table(Month = rep(200004, 5), Firm = rep(3, 5), Rt = rnorm(5)),
data.table(Month = rep(200005, 2), Firm = rep(3, 2), Rt = rnorm(2)))
DT <- rbind(Firm1, Firm2, Firm3)
DT[, Mar := rnorm(50)]
哪个给出了
Month Firm Rt Mar
200001 1 -1.34767475 0.865598407
200001 1 -0.70741105 -0.782668556
200001 1 0.61342578 0.021440129
200002 1 -1.53156217 1.988291260
200002 1 -0.42512876 -0.384017585
...
在此示例中，多个“月”因素与每日观察有关。每个公司每个月都有不同数量的每日观察结果。
我想做的是从Rt与Mar的线性回归估计残差的方差。为了每月这样做，我会使用
DT[, var(lm(Rt ~ Mar)$residuals), by = c("Firm", "Month")]
与上述不同，我希望每个月都使用之前的n个月的观察结果。天真的尝试可能看起来像
DT[, var(lm(Rt ~ Mar)$residuals), by = c("Firm", Month[t : t-2])]
最终输出
Month Firm Rt Mar resVariance
200001 1 -1.34767475 0.865598407 NA
200001 1 -0.70741105 -0.782668556 NA
200001 1 0.61342578 0.021440129 NA
200002 1 -1.53156217 1.988291260 NA
200002 1 -0.42512876 -0.384017585 NA
200002 1 1.06399050 -1.123293332 NA
200002 1 -1.59751358 -0.188190495 NA
200003 1 0.08854875 0.897471055 0.8745559
200003 1 0.37822085 -0.654418019 0.8745559
200003 1 1.07786336 1.665720591 0.8745559
200003 1 0.92820233 -1.983931767 0.8745559
200003 1 -0.43148095 -0.286302699 0.8745559
200004 1 -0.80384703 -0.927657523 0.841801
200004 1 -0.79383439 1.281582524 0.841801
...
对此的任何建议都将不胜感激。