四分之一假人的线性回归

Question

I am trying to fit a linear regression to the data below

Power<-mutate(Power,Year=format(Date,"%Y"),Quarter=quarters(Date),Month=format(Date,"%m"))
head(Power)
       Date    YY    XX  Year    Quarter
2007-01-01     NA     NA 2007      Q1
2007-01-02     NA     NA 2007      Q1
2007-01-03  55.90  71.40 2007      Q1
2007-01-04  55.25  70.75 2007      Q1

The model is

lm(YY~XX+as.factor(Quarter,ref="Q1"),data=Power)

This works fine. However, it automatically creates three dummies for 3 quarters. Is there any way to include only one dummy, say Q2 in this model?

Answer 1

Arguably the most common way to do this is to create a dichotomous variable on the fly with I() .

lm(YY ~ XX + I(Quarter=="Q2"), data=Power)

This includes a binary predictor in the model that's 1 when Quarter=="Q2" and 0 otherwise.

Answer 2

Easiest way would be to create new variable with information you need...

Power$Q2dummy <- 0
Power$Q2dummy[which(Power$Quarter == 'Q2')] <- 1
lm(YY~XX+Q2dummy,data=Power)

However, it is hard to say, because you dont provide your data or even their summary (what is variable Quarter? Factor with 4 states I guess?).

Answer 3

One possibility could be to use the ifelse(rule, if TRUE, if FALSE) command :

For Example:

Power$Q2dummy <- ifelse(Power$Quarter == "Q2",1,0)
lm(YY~XX+Q2dummy,data=Power)