问题
So I want to plot this:
lmfit = lm (y ~ a + b)
but, "b" only has the values of zero and one. So, I want to plot two separate regression lines, that are paralel to one one another to show the difference that b makes to the y-intercept. So after plotting this:
plot(b,y)
I want to then use abline(lmfit,col="red",lwd=2) twice, once with the x value of b set to zero, and once with it set to one. So once without the term included, and once where b is just 1b.
To restate: b is categorical, 0 or 1. a is continuous with a slight linear trend.
Thank you.
Example:
回答1:
You might want to consider using predict(...) with b=0 and b=1, as follows. Since you didn't provide any data, I'm using the built-in mtcars dataset.
lmfit <- lm(mpg~wt+cyl,mtcars)
plot(mpg~wt,mtcars,col=mtcars$cyl,pch=20)
curve(predict(lmfit,newdata=data.frame(wt=x,cyl=4)),col=4,add=T)
curve(predict(lmfit,newdata=data.frame(wt=x,cyl=6)),col=6,add=T)
curve(predict(lmfit,newdata=data.frame(wt=x,cyl=8)),col=8,add=T)
回答2:
Given you have an additive lm model to begin with, drawing the lines is pretty straightforward, even though not completely intuitive. I tested it with the following simulated data:
y <- rnorm(30)
a <- rep(1:10,times=3)
b <- rep(c(1,0),each=15)
LM <- lm(y~a+b)
You have to access the coefficient values in the lm. Its is:
LM$coefficients
Here comes the tricky part, you have to assign the coefficients for each line.
The first one is easy:
abline(LM$coef[1],LM$coef[2])
The other one is a bit more complicated, given R works with additive coefficients, so for the second line you have:
abline(LM$coef[1]+LM$coef[3],LM$coef[2])
I hope this is what you was expecting
回答3:
Unless I've misunderstood the question, all you have to do is run abline again but on a model without the b term.
abline(lm(y~a),col="red",lwd=2)
来源：https://stackoverflow.com/questions/26895778/how-to-plot-a-comparisson-of-two-fixed-categorical-values-for-linear-regression