How to choose variable pairs for regression analysis using lapply

Question

I have a following dataframe:

varnames<-c( "id", "a", "b", "c", "a.0", "b.0", "c.0", "d", "e", "f")
a <-as.data.frame (matrix (round (rnorm (40, mean = 5, sd = 3), 1), 4, 10))
colnames (a)<-varnames

> a
   id    a   b   c a.0 b.0 c.0   d    e   f
1 2.1 -0.5 7.0 6.0 1.7 9.2 2.6 3.6  4.6 7.0
2 4.0  7.1 4.1 5.4 6.8 4.3 7.7 0.1  5.6 1.4
3 8.7  6.9 3.8 7.5 5.1 3.3 7.9 4.1 10.3 5.7
4 5.6  5.3 3.9 2.8 8.2 4.9 6.9 6.6  3.7 1.9

I am performing three separate GEE analyses with formula always containing "d", "e", and "f" as predictor variables (d+e+f). Each formula has either "a" or "b" or "c" as dependent variable:

library (gee)
model<-lapply ((a)[2:4], function (x) {
gee(x~d+e+f, data=a, id = id)
}) 

This code works OK.

My next aim is to add "a.0" or "b.0" or "c.0" as predictor variables to formulas having "a", "b" and "c" as dependent variables, correspondingly:

gee(a~a.0+d+e+f, data=a, id = id)

gee(b~b.0+d+e+f, data=a, id = id)

gee(c~c.0+d+e+f, data=a, id = id)

That is, I would like adding to each formula an additional predictor variable corresponding to dependent variable, but having ".0" suffix.

I naively tried the following without much success:

model<-lapply ((a)[2:4], function (x) {
gee(x~ paste(names (x)[2:4], "0", sep = ".")+d+e+f, data=a, id = id)
})

It probably does not work as code " paste(names (x)[2:4], "0", sep = ".") " within function does not return proper variable name, although it does so when used as a separate code.

I would be most grateful to your suggestions on how to automatically insert proper variable names in my formulas.

Answer 1

Instead of working on the columns of a, you could operate on the names of the variables you want to process:

set.seed(144)
varnames<-c( "id", "a", "b", "c", "a.0", "b.0", "c.0", "d", "e", "f")
a <- as.data.frame (matrix (round (rnorm (40, mean = 5, sd = 3), 1), 4, 10))
colnames(a) <- varnames
model <- lapply(varnames[2:4], function(x) {
  gee(paste0(x, "~", x, ".0+d+e+f"), data=a, id=id)
})