In linear regression, regression weights of x1 on Y given x2 and x3 should
be mathematically identical to the semipartial correlations between x1 and
Y, given x2 and x3.
However, I do not obtain identical results, so apparently I'm doing
something wrong in R.
Data preparation:
data<-read.csv("file.csv", head=T)
data <- subset(data, select=c(age,s1,s2,s3,s5,s10,WSAS01))
data <- na.omit(data1)
In the following example, I am interested in the regression
coefficient/semi-partial correlation of s1(x) and WSAS01 (y).
Regression:
m1 <- lm(WSAS01 ~ s1+s2+s3+s5+s10+age, data=data)
summary(m1)
regression coefficient s1: 0.091
Semipartial Correlation:
library(ppcor)
spcor.test(data$s1,data$WSAS01,data[,c("age","s2","s3","s5","s10")])
semi-partial correlation s1 with WSAS01: 0.202
What am I doing wrong?
Thank you
T
[[alternative HTML version deleted]]
