面板回归误差

Question

我正在尝试进行面板回归，其中因变量（各种公司的股票回报）对 5 个自变量进行回归。 这是自变量数据框的可复制示例

dput(factors_1[1:10,])
structure(list(Date = 200002:200011, Mkt.RF = c(5.94, 0.66, -5.58, 
-0.09, 0.67, -1.58, -1.61, -4.99, -2.71, -4.55), SMB = c(0.84, 
-5.15, -4.62, 0.16, 0.33, -0.69, 0.68, 2.35, -6.1, -0.78), HML = c(-9.45, 
3.33, 5.93, 6.17, 3.14, 3.31, -0.5, 2.64, 7.54, 11.15), RMW = c(3.55, 
-2.59, -1.53, -3.38, -3.45, -0.12, -1.27, 1.63, 2.7, 0.79), CMA = c(-7.33, 
4.96, 1.32, 4.94, 1.22, -0.12, 0.64, 2.16, 4.1, 8.75), RF = c(0.43, 
0.47, 0.46, 0.5, 0.4, 0.48, 0.5, 0.51, 0.56, 0.51)), row.names = c(NA, 
10L), class = "data.frame")

这里是股票回报

dput(xx[1:10, 1:10])
structure(list(Date = structure(c(10990.9954886386, 11019.9953776753, 
11050.9954014418, 11080.9952984982, 11111.9953776753, 11141.9951640545, 
11172.995061378, 11203.9951324494, 11233.9950455918, 11264.9949982497
), class = "Date"), X1 = c(0.0954887438827963, 
-0.0596463008222008, 0.071350885788402, 0.0305926490738153, 0.0408331711459304, 
-0.0211402933162625, -0.00493862203119688, 0.006182173191563, 
0.0032423131269943, 0.0193884936176278), X2 = c(-0.123462974283698, 
0.230503533400868, -0.0272942506612435, 0.0483790669291113, -0.0595278152717571, 
0.12087834022411, -0.032011380068422, -0.0813892896957779, 0.0138779835292666, 
0.0726322608057619), X3 = c(-0.0682052985267971, 0.172249290323711, 
-0.154888201350603, 0.0395159403332963, -0.0143942598523314, 
-0.0607566985291722, -0.0310708779173386, -0.0746345858888015, 
-0.151109426840925, 0.0118888362760825), X4 = c(-0.114511361380472, 
0.00998441685033158, 0.192522150537581, -0.0158023343537101, 
0.0374730915541921, 0.0777493327863055, -0.0016218724457906, 
-0.0635452365157563, 0.0565030106039299, 0.115759209185826), 
   X5 = c(0.00389199996406542, -0.0212889913893688, 
    0.164892967212694, -0.00832469019706505, -0.00462232472270219, 
    -0.0070177636719938, 0.00453659662769512, 0.0528941822866427, 
    -0.00836737746775751, -0.0050017502848112), X6 = c(-0.10351479457366, 
    0.0237125822002096, 0.0101860439504515, 0.0111924296807739, 
    -0.0652473862813747, 2.11404059631271e-05, 0.0261396151198399, 
    -0.0356789492292369, -0.0706069184275196, -0.0656535040135704
    ), X7 = c(-0.0980023956049211, 0.102552120231041, 
    -0.0959174074104425, -0.0790740833989735, 0.118610740782993, 
    -0.100050822390369, -0.00333557692764708, -0.0368703292701125, 
    0.0628135821343774, 0.0471186471744018), X9 = c(-0.0304322345046196, 
    -0.0977595796246631, 0.138258584646108, 0.0344876873979214, 
    -0.000721154371596811, 0.0508635363751093, 0.0533435865577603, 
    -0.0506646520146184, 0.0497235991059199, 0.0284083879640369
    ), X9 = c(-0.159712703662352, -0.0234902492510041, 0.116858931667507, 
    0.00432376896685471, 0.114340108193219, 0.00235829911414087, 
    -0.0573195744121493, 0.095634961997471, -0.0871461890063988, 
    -0.0738243041819919)), row.names = c(NA, 10L), class = "data.frame")

我尝试了什么：

p1_q1_l<-plm(as.matrix(data.frame(xx[, -1]))~factors_1$Mkt.RF+factors_1$SMB+factors_1$HML+factors_1$RMW+factors_1$CMA,data=factors_1, method="within")

我得到了什么

Error in tapply(x, effect, func, ...) : arguments must have same length

我不明白发生了什么。两个表都是具有相同观察数的数据框。我该如何解决这个问题？

Answer 1

错误很可能是由于您将矩阵定义为独立 (Y) 变量，其中需要向量。您需要长格式的数据，其中 Y 是 one 列，ID 和 time 列表示不同的观察结果。

不过，我对您的两个数据集的兼容性有一些疑问，但您可能希望将它们合并为一个。仔细看看，如何合并原始数据，尤其是关于 Date 列。

当我正确理解您的xxx 数据时，X* 是不同的公司。现在使用reshape 将数据转换为长格式。

xxx <- reshape(xx, timevar=1, varying=2:9, direction="long", sep="")
xxx$Date <- as.character(xx$Date[xxx$Date])

然后，将merge 两个数据集合二为一可能会更容易。但是，两个数据帧的 "Date" 列不匹配。当我正确理解您的factors_1 数据时，它们是每月值。我将继续简单地将"03" 附加到第二个以使它们现在匹配，但是您知道实际需要什么。

factors_1x <- transform(factors_1, 
                        Date=as.character(as.Date(strptime(paste(factors_1$Date, 03),
                                                           "%Y%m%d"))))

这里合并。

dat <- merge(xxx, factors_1x, all.x=TRUE)
head(dat)
#         Date           X id Mkt.RF  SMB   HML  RMW   CMA   RF
# 1 2000-02-03  0.09548874  1   5.94 0.84 -9.45 3.55 -7.33 0.43
# 2 2000-02-03 -0.05964630  2   5.94 0.84 -9.45 3.55 -7.33 0.43
# 3 2000-02-03  0.07135089  3   5.94 0.84 -9.45 3.55 -7.33 0.43
# 4 2000-02-03  0.03059265  4   5.94 0.84 -9.45 3.55 -7.33 0.43
# 5 2000-02-03  0.04083317  5   5.94 0.84 -9.45 3.55 -7.33 0.43
# 6 2000-02-03 -0.02114029  6   5.94 0.84 -9.45 3.55 -7.33 0.43

现在更容易编写公式了。可以使用index=c("id", "Date") 在plm 调用中制定新索引。

library(plm)
p1_q1_l <- plm(X ~ Mkt.RF + SMB + HML + RMW + CMA + RF, method="within",
               index=c("id", "Date"), data=dat)
# Model Formula: X ~ Mkt.RF + SMB + HML + RMW + CMA + RF
# 
# Coefficients:
#     Mkt.RF        SMB        HML        RMW        CMA         RF 
#  0.0042267  0.0054278 -0.0016806  0.0129446  0.0148160 -0.4194726