Some ideas about grammar of statistics language


Many problems, such as some in Philosophy, are difficult because it is hard to state them clearly. A powerful language system improves people’s ability to solve complicated problems.

R is almost a statistics language. Sometimes it helps to think in R commands. For example, pnorm(2.5, 2, 4), qt(.975, 20) is more intuitive than mathematical notations. And by, *ply functions will help to think about matrix computations.

It is a good idea to melt programming language and mathematical language together. Sentences, or code can help to understand and can be run at the same time. This requires a programming language to be very expressive. R is far from enough.

I want R to support dat[x1==1, ]. It’s better than dat[dat$x1==1, ]. And also I hope I can use | as conditioning, for example mean(y|x) where x is a category vector. I don’t know how to implement this two functions in R.

There is no sum of squares function in R. In linear model, this is the most basic function. So I write a few piece of codes to make life easier, such as functions to calculate MS(y|x) = [E(y|x) - E(y)] ^ 2.

In the following code, MS(y|x) is ss(project(y, x)) / (nleves(x) - 1).

#compute sum of squares
ss=function(x, na.rm=FALSE){
    sum((x - mean(x)) ^ 2, na.rm=na.rm)

# Compute the projection of y given x. Currently y can only be numeric.
# Note: When calculating the ss for interaction terms, we need to remove the contribution of linear parts.

project=function(y, x){
        print("sorry, method for project factors is still under developing")
        x <- as.factor(x)
        x=interaction(x, sep="")
        projected.values <- tapply(y, x, mean)
        projection <- x
        levels(projection) <- projected.values
        projection <- as.numeric(as.character(projection))
        x = as.matrix(x)
        y = as.matrix(y)
        return(x %*% solve(crossprod(x)) %*% t(x) %*% y)