Accessing Regression Model Variables in R
I am working with a linear model, say
y<-rnorm(20) x1<-rgamma(20,2,1) x2<-rpois(20,3) fit<-lm(y~x1*x2) summary(fit)
and I was wondering if there is a way to access the regression variables via lm
? One option would be to just use
fit$model
and what do you get
y x1 x2
1 1.52366782 1.1741392 4
2 -0.23640711 0.8780224 2
3 0.90728028 3.2192029 3
4 -0.07964816 2.5476499 3
5 -0.50999322 1.8515032 2
6 0.08854942 0.2892199 1
7 0.19708785 1.1865428 3
8 0.09641827 0.5808471 1
9 -0.28815596 1.3589239 1
10 -0.45784790 3.7514056 2
11 -0.39785151 0.8648221 0
12 -0.17503763 0.8153766 3
13 1.44095562 0.1933351 1
14 -0.86787846 2.7348324 5
15 0.30369142 0.7547339 7
16 -0.76884945 2.1558952 1
17 -0.81620973 1.2373447 3
18 -0.40978079 1.2046777 2
19 -0.23160424 1.8455335 2
20 2.90504457 0.9968425 2
and then the answer is displayed along with the covariates. The problem is that x1:x2
there is no interaction , and this is often an important part of model building.
The problem arises in the context of variable selection, where step-by-step procedures reduce a huge dataset to a few variables and their interactions. I would like to plot a covariance matrix for this model without reference to the dataset.
Thanks in advance.
source to share
Thanks Roland
You can get the model string by accessing the summary attributes (take a look str(s)
)
s <- summary(fit)
mymod <- paste(attr(s$terms, "term.labels"), collapse=" + ")
mymod
[1] "x1 + x2 + x1:x2"
However, you can get the data by passing in a model suitable for model.matrix
model.matrix(fit)
There will be a little extra work if you have variables factor
.
source to share