在R中使用as.numeric（）转换时出错

Question

I have a dataset:

 > x
    Treatment X1 X2
1         T1  6  7
2         T1  5  9
3         T1  8  6
4         T1  4  9
5         T1  7  9
6         T2  3  3
7         T2  1  6
8         T2  2  3
9         T3  2  3
10        T3  5  1
11        T3  3  1
12        T3  2  3

I am trying to find means of the columns X1 and X2. If I run the data as-is, I get an error:

> t1 <- subset(x[2:3], x$Treatment=="T1")
> x_vec <- colMeans(t1, na.rm = TRUE)
Error in colMeans(t1, na.rm = TRUE) : 'x' must be numeric

So, I need to convert X1 and X2 to numeric:

t1$X1 <- as.numeric(as.factor(t1$X1))
t1$X2 <- as.numeric(as.factor(t1$X2))
x_vec <- colMeans(t1, na.rm = TRUE)

But when I do that, I get the wrong result:

> x_vec
 X1  X2 
6.0 4.4 

The t1, after conversion to as.numeric() , shows:

> t1
  X1 X2
1  6  4
2  5  5
3  8  3
4  4  5
5  7  5

Why are the values in X2 changed after converting to numeric?

Answer 1

This is a pretty common issue that newer R users hit. The issue is your use of as.factor . running as.numeric on a factor converts the value to the numeric index of the label, rather than converting the label itself to a number. Your can either remove the call to as.factor or run as.character on the factor before calling as.numeric .

Note that some functions like as.data.frame automatically convert characters to factors, which can cause problems. Check out the option stringsAsFactors for more info.