Question

我正在尝试执行以下操作：

使用移动的24个月窗口拟合逐步回归模型
在每个步骤中，预测基于模型的退货1个月
计算每个预测的标准误差
绘制实际回报，预测回报，标准错误

我的数据如下（第一个是依赖，第二个和第三个是自变量） - 我无法让格式看起来更好，并且目前无法使用dropbox来附加文件：

FundA.Return    Explanatory.Variable 1  Explanatory.Variable 2
2.17%       1.96%           -2.34%
0.35%       -2.11%          -2.00%
1.18%       0.29%           -0.09%
2.23%       0.47%           1.35%
2.38%       2.04%           3.73%
1.39%       -0.39%          -1.54%
-0.63%      -0.16%          -4.20%
-2.04%      -0.18%          0.53%
1.80%       0.23%           -1.85%
2.19%       0.66%           1.10%
-0.22%      -1.59%          3.68%
0.68%       -0.12%          -1.96%
0.75%       -0.27%          -1.76%
3.41%       0.52%           -0.42%
-1.18%      -0.75%          3.72%
-0.17%      0.36%           3.49%
1.98%       0.85%           1.77%
1.21%       -0.33%          4.04%
-1.38%      0.17%           -3.73%
-0.87%      -0.37%          0.27%
-4.40%      2.20%           -0.75%
-1.18%      6.31%           -0.56%
0.43%       -1.28%          -2.12%
0.24%       3.11%           -2.58%
2.27%       0.15%           6.85%
0.36%       0.49%           1.94%
0.91%       -0.38%          -1.18%
1.69%       1.18%           3.21%
3.17%       3.39%           3.06%
1.46%       3.93%           -1.74%

我的代码如下（稍微编辑但我仍然收到错误）：

library(ISLR)
Returns <- read.csv("Returns New Factors2.csv",header=TRUE)
library(leaps)
P <- nrow(Returns)-24
results <- rep(0,P)
for(i in 1:P){
  stepfit <- regsubsets(FundA.Return~.,data=Returns[i:24+i,],method="forward")
  coefi <- coef(stepfit,id=1)
  pred = model.matrix[P-24-i+1,]%*%coefi
  summary(coefi)
  SE[i+1] = ((Returns$FundA.Return[i+1]-pred)^2)
  plot(FundA.Return,pred,SE)
}

我收到以下错误。我做错了什么？

model.matrix中的错误[P - 24 - i + 1，]： 'closure'类型的对象不是可子集化的

移动窗口逐步回归预测

0 个答案: