为什么内置的lm函数在R中这么慢?

我一直认为R中的lm函数非常快，但是正如本示例所建议的那样，使用solve函数计算的封闭解更快.

I always thought that the lm function was extremely fast in R, but as this example would suggest, the closed solution computed using the solve function is way faster.

data<-data.frame(y=rnorm(1000),x1=rnorm(1000),x2=rnorm(1000))
X = cbind(1,data$x1,data$x2)

library(microbenchmark)
microbenchmark(
solve(t(X) %*% X, t(X) %*% data$y),
lm(y ~ .,data=data))

如果这个玩具示例是一个不好的示例，或者是lm实际上很慢的情况，有人可以向我解释吗?

Can someone explain me if this toy example is a bad example or it is the case that lm is actually slow?

根据Dirk Eddelbuettel的建议，由于lm需要解析公式，所以比较是不公平的，因此最好使用lm.fit而不需要解析公式

As suggested by Dirk Eddelbuettel, as lm needs to resolve the formula, the comparison is unfair, so better to use lm.fit which doesn't need to resolve the formula

microbenchmark(
solve(t(X) %*% X, t(X) %*% data$y),
lm.fit(X,data$y))


Unit: microseconds
                           expr     min      lq     mean   median       uq     max neval cld
solve(t(X) %*% X, t(X) %*% data$y)  99.083 108.754 125.1398 118.0305 131.2545 236.060   100  a 
                      lm.fit(X, y) 125.136 136.978 151.4656 143.4915 156.7155 262.114   100   b