my first R question that has not been discussed in any forum yet, apparently...Consider my fake dataset:
A<-matrix(c(1,2,3,4,5,2,3,4,5,6,3,4,5,6,7),5,3)
a<-c(2,4,6,8,9)
I want to regress each column of A on a and perform systemfit to test some restrictions, e.g.:
system.1<-list(A[,1]~a,A[,2]~a,A[,3]~a)
systemfit(system.1)
Now my problem is that my "real" matrix A has hundreds of columns. I'm struggling to create a list that systemfit accepts. I've come up with the following, not workable code:
varlist=NULL
for (i in 1:3){varlist[i] <- paste("A[,",i,"] ~ a",sep="")}
models <- lapply(varlist, function(x){
systemfit(substitute(j, list(j = as.name(x))))
})
If you hit
substitute(j, list(j = as.name(varlist)))
you can see that the solution
`A[,1] ~ a`
contains `` signs which seem to be causing the trouble for systemfit, since it is not accepted as a formula. Hence the problem seems to be the columnwise looping, but I dont see any alternative for the dataset at hand...Any ideas?
Any help would be highly appreciated!
Thanks!
The idiomatic way to do this is to create a list of formulas which reference columns in a data frame, then pass the list and the data frame to systemfit(...).
df <- data.frame(a,A) # data frame with columns a, X1, X2, X3, ...
forms <- lapply(paste0(colnames(df)[-1],"~a"),as.formula)
library(systemfit)
systemfit(forms,data=df)
# systemfit results
# method: OLS
#
# Coefficients:
# eq1_(Intercept) eq1_a eq2_(Intercept) eq2_a eq3_(Intercept) eq3_a
# -0.182927 0.548780 0.817073 0.548780 1.817073 0.548780