How to create a linear model with interaction term only in R?

R ProgrammingServer Side ProgrammingProgramming

To create a linear model with interaction term only, we can use the interaction variable while creating the model. For example, if we have a data frame called df that has two independent variables say V1 and V2 and one dependent variable Y then the linear model with interaction term only can be created as lm(Y~V1:V2,data=df).

Consider the below data frame −

Example

 Live Demo

x1<-rnorm(20,5,1.2)
x2<-rnorm(20,2,1.2)
y1<-rnorm(20,3,1.25)
df1<-data.frame(x1,x2,y1)
df1

Output

     x1       x2         y1
1 4.442636  2.63714281  0.9120181
2 5.912804  1.72409663  2.2746513
3 7.881736  3.48780844  3.2675991
4 6.145395  1.51225991  3.2177194
5 6.303181  4.34733534  3.3361897
6 5.689431  0.08564247  2.0863346
7 6.008397  1.30221198  2.7183725
8 6.580716  1.45239970  4.0301650
9 4.000381  3.71539569  3.9404225
10 5.535338  3.48976926  4.9557343
11 5.968661  2.18208048  3.9992293
12 4.783326  -0.90756376 4.2661775
13 3.843528  -0.50299042 0.6496255
14 6.034803  1.06019093  5.2339701
15 4.177977  1.70476660  2.5932408
16 5.746300  0.85021792  3.0145627
17 4.658000  1.49894326  5.2826538
18 3.797983  0.46900834  3.6511403
19 5.415130  -0.31811078  2.6574529
20 4.709443  2.72001915  1.8893078

Creating linear model for y1 with interaction between x1 and x2 −

Model1<-lm(y1~x1:x2,data=df1)
summary(Model1)

Call −

lm(formula = y1 ~ x1:x2, data = df1)

Residuals −

Min 1Q Median 3Q Max
-2.3517 -0.6387 -0.1708 0.7321 2.1390

Coefficients −

Estimate Std. Error t value Pr(>|t|)
(Intercept) 2.96669 0.42803 6.931 1.77e-06 ***
x1:x2 0.02535 0.03434 0.738 0.47
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1


Residual standard error: 1.299 on 18 degrees of freedom
Multiple R-squared: 0.02939, Adjusted R-squared: -0.02454
F-statistic: 0.545 on 1 and 18 DF, p-value: 0.4699

Example

 Live Demo

v1<-rpois(20,2)
v2<-rpois(20,2)
v3<-rpois(20,2)
Response<-rpois(20,5)
df2<-data.frame(v1,v2,v3,Response)
df2

Output

  v1 v2 v3 Response
1 2  2  4  10
2 1  1  2  2
3 3  4  3  4
4 2  1  2  5
5 0  0  2  4
6 0  2  1  5
7 1  3  0  5
8 0  0  1  6
9 1  3  1  6
10 3 0  2  6
11 2 1  0  3
12 1 3  1  2
13 2 1  3  5
14 0 1  0  8
15 0 3  1  4
16 1 0  3  2
17 3 1  3  2
18 0 4  1  9
19 1 2  3  9
20 3 2  0  9

Creating linear model for Response with interaction between v1, v2, and v3 −

Model2<-lm(Response~v1:v2+v2:v3+v1:v3+v1:v2:v3,data=df2)
summary(Model2)

Call −

lm(formula = Response ~ v1:v2 + v2:v3 + v1:v3 + v1:v2:v3, data = df2)

Residuals −

Min 1Q Median 3Q Max
-4.3677 -1.7339 -0.0289 1.7127 4.0160

Coefficients −

Estimate Std. Error t value Pr(>|t|)
(Intercept) 3.98404 1.21641 3.275 0.00511 **
v1:v2 0.31934 0.44221 0.722 0.48131
v2:v3 0.83716 0.46012 1.819 0.08886 .
v1:v3 0.02783 0.30125 0.092 0.92762
v1:v2:v3 -0.37123 0.28280 -1.313 0.20901
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1


Residual standard error: 2.609 on 15 degrees of freedom
Multiple R-squared: 0.1907, Adjusted R-squared: -0.02517
F-statistic: 0.8834 on 4 and 15 DF, p-value: 0.4973
raja
Published on 06-Feb-2021 11:31:49
Advertisements