我想有四个预测我在哪里可以自由指定用固定R2模拟多元回归数据:如何合并相关变量?
- 整体解释模型
- 所有标准化回归系数的大小的变化来模拟数据进行多元线性回归
- 的预测变量彼此相关的程度
我到达了满足前两点的解决方案,但是基于所有ind自变量彼此无关(请参阅下面的代码)。为了得到标准化的回归系数,我从平均值= 0和方差= 1的总体变量中抽样。
# Specify population variance/covariance of four predictor variables that is sampled from
sigma.1 <- matrix(c(1,0,0,0,
0,1,0,0,
0,0,1,0,
0,0,0,1),nrow=4,ncol=4)
# Specify population means of four predictor varialbes that is sampled from
mu.1 <- rep(0,4)
# Specify sample size, true regression coefficients, and explained variance
n.obs <- 50000 # to avoid sampling error problems
intercept <- 0.5
beta <- c(0.4, 0.3, 0.25, 0.25)
r2 <- 0.30
# Create sample with four predictor variables
library(MASS)
sample1 <- as.data.frame(mvrnorm(n = n.obs, mu.1, sigma.1, empirical=FALSE))
# Add error variable based on desired r2
var.epsilon <- (beta[1]^2+beta[2]^2+beta[3]^2+beta[4]^2)*((1 - r2)/r2)
sample1$epsilon <- rnorm(n.obs, sd=sqrt(var.epsilon))
# Add y variable based on true coefficients and desired r2
sample1$y <- intercept + beta[1]*sample1$V1 + beta[2]*sample1$V2 +
beta[3]*sample1$V3 + beta[4]*sample1$V4 + sample1$epsilon
# Inspect model
summary(lm(y~V1+V2+V3+V4, data=sample1))
Call:
lm(formula = y ~ V1 + V2 + V3 + V4, data = sample1)
Residuals:
Min 1Q Median 3Q Max
-4.0564 -0.6310 -0.0048 0.6339 3.7119
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.496063 0.004175 118.82 <2e-16 ***
V1 0.402588 0.004189 96.11 <2e-16 ***
V2 0.291636 0.004178 69.81 <2e-16 ***
V3 0.247347 0.004171 59.30 <2e-16 ***
V4 0.253810 0.004175 60.79 <2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 0.9335 on 49995 degrees of freedom
Multiple R-squared: 0.299, Adjusted R-squared: 0.299
F-statistic: 5332 on 4 and 49995 DF, p-value: < 2.2e-16
问题:如果我的预测变量是相关的,因此,如果没有非对角线元素为0指定其方差/协方差矩阵中,R2和回归系数主要来自我多么希望他们能有所不同,例如通过使用
sigma.1 <- matrix(c(1,0.25,0.25,0.25,
0.25,1,0.25,0.25,
0.25,0.25,1,0.25,
0.25,0.25,0.25,1),nrow=4,ncol=4)
有什么建议吗? 谢谢!