Replacing '+' of a formula by ',' and list them with quotation marks

[lherming]

I try to combine an output in a different way, so that in the end I get the columns values of selected variables after the backward selection. Unfortunately I have problems with the combination of the variables. After the backward selection the variables are output as formula and I would have liked to separate this formula by , instead of + and put the variables in quotation marks "". I tried different ways and unfortunately I did not get a result.

external <- ts(mtcars, start = c(2015,1), end = c(2019,52), frequency = 52)
# define all possible variables for variable selection
x <- window(external, end= c(2018,52))
data.variables <- lm(mpg ~cyl+disp+hp+drat+wt+qsec+vs+am+gear+carb, data=window(external, start=c(2015,1), end=c(2018,52)))
# Variable selection backwards
step <- step(data.variables, direction="backward", k=log(208))  
# provide matrix regression
## Here I want to paste the calculated Variables listed like this: "disp","hp","wt","qsec","am"
xregs <- as.matrix(window(external[,c(paste(gsub("+", ",",(deparse(step[["call"]][["formula"]][[3]])), fixed=T))), end = c(2018, 52))) 
# the acutal line should look like this:
xregs <- as.matrix(window(external[,c("disp","hp","wt","qsec","am")], end = c(2018, 52)))
# but I only receive this:
xregs <- as.matrix(window(external[,c("disp,hp,wt,qsec,am")], end = c(2018, 52)))

Especially when I use the following:

a <-gsub("+", ",",(deparse(step[["call"]][["formula"]][[3]])), fixed=T)
cat(sapply(strsplit(a, '[, ]+'), function(x) toString(dQuote(x, FALSE))))
#"disp", "hp", "wt", "qsec", "am"

the right line is printed, but I don't know how to use the printed line in the other command.

[woodward]

How about this?

library(stringr)
vs <- deparse(step[["call"]][["formula"]][[3]]) # get formula from lm
vs <- str_trim(str_split(vs, pattern = "\\+", simplify = TRUE)) # split string
xregs <- as.matrix(window(external[,vs], end = c(2018, 52))) # subset x

[lherming]

Thank you first of all for your help. It does work!

But if i apply this to my data, where way more variable are chosen, I actually get:

My procedure and my results:

  # provide xregs for training data
  step <- step(data.variables, direction="backward", k=log(length(x)))    
  # provide xregs for training data
  vs <- deparse(step[["call"]][["formula"]][[3]]) # get formula from lm
#[1] "F7_1 + F11_1 + F32_1 + F33_0 + F34__1 + F37_1 + F39_2 + F61_0 + "
#[2] "    AO12 + LS39 + AO63 + AO71"   
  a <- str_trim(str_split(vs, pattern = "\\+", simplify = TRUE)) # split string
# [1] "F7_1"   "AO12"   "F11_1"  "LS39"   "F32_1"  "AO63"   "F33_0"  "AO71"   "F34__1" ""      
#[11] "F37_1"  ""       "F39_2"  ""       "F61_0"  ""       ""       ""     
  xregs <- as.matrix(window(external_data.ccf_ts[,a], end = c(2018, 52)+(i-1)/52)) # subset x
#Error in `[.default`(external_data.ccf_ts, , a) : 
#Indexing outside the boundaries

I think it cant handle two rows of formula. Do you have any idea how to solve this?

[woodward]

You need to add something like this to combine the rows of the formula.

vs <- paste(vs)

[nirgrahamuk]

@lherming

step <- step(data.variables, direction="backward", k=log(208))

some advice, for your own sanity, try to avoid naming objects you make with the name of existing R functions that you use. with that in mind , assuming step object is called step_ to differentiate it from the step function then you could do

attr(step_$terms,"term.labels") %>% paste0(collapse=",")
#"disp,hp,wt,qsec,am"

instead of the deparse stuff to get the indepentent terms out as a character string with comma sep