In this experiment, four different diets were tried on animals. Then researchers measured their effects on blood coagulation time.
## Data :
coag diet
1 62 A
2 60 A
3 63 A
4 59 A
5 63 B
6 67 B
7 71 B
8 64 B
9 65 B
10 66 B
11 68 C
12 66 C
13 71 C
14 67 C
15 68 C
16 68 C
17 56 D
18 62 D
19 60 D
20 61 D
21 63 D
22 64 D
23 63 D
24 59 D
I am trying to fit a linear model for coag~diet by using the function lm in R Results should look like the following:
> modelSummary$coefficients
Estimate Std. Error t value Pr(>|t|)
(Intercept) 6.100000e+01 1.183216 5.155441e+01 9.547815e-23
dietB 5.000000e+00 1.527525 3.273268e+00 3.802505e-03
dietC 7.000000e+00 1.527525 4.582576e+00 1.805132e-04
dietD -1.071287e-14 1.449138 -7.392579e-15 1.000000e+00
My code thus far does not look like results:
coagulation$x1 <- 1*(coagulation$diet=="B")
coagulation$x2 <- 1*(coagulation$diet=="C")
coagulation$x3 <- 1*(coagulation$diet=="D")
modelSummary <- lm(coag~1+x1+x2+x3, data=coagulation)
"diet" is a character variable and is treated as a factor. So you may leave out the dummy coding and just do:
summary(lm(coag ~ diet, data=coagulation))$coefficients
# Estimate Std. Error t value Pr(>|t|)
# (Intercept) 6.100000e+01 1.183216 5.155441e+01 9.547815e-23
# dietB 5.000000e+00 1.527525 3.273268e+00 3.802505e-03
# dietC 7.000000e+00 1.527525 4.582576e+00 1.805132e-04
# dietD 2.991428e-15 1.449138 2.064281e-15 1.000000e+00
Even if "diet" were a numeric variable and you want R to treat it as a categorical rather than a continuous variable no dummy coding is needed, you would just add it as + factor(diet) into the formula.
As you see, also 1 + is redundant since lm calculates the (Intercept) by default. To omit the intercept, you may do 0 + (or - 1).