Given:
vars <- c("var1", "var2", "var3", "var4")
mm_exp <- expression(
f(testm, testmodel = 'test', testgraph = g, truetest1 = TRUE, truetest2 = TRUE),
f(testm2, testmodel = 'fg'),
f(testm3, testmodel = 'fg3')
)
I want to produce formulas using mm_exp and all combinations (combn) of vars to enter in a model:
#y ~ var1 + var2 + var3 + var4 + f(testm, testmodel = "test", testgraph = g, truetest1 = TRUE,
# truetest2 = TRUE) + f(testm2, testmodel = "fg") + f(testm3, testmodel = 'fg3')
#y ~ var1 + var2 + var3 + f(testm, testmodel = "test", testgraph = g, truetest1 = TRUE,
# truetest2 = TRUE) + f(testm2, testmodel = "fg") + f(testm3, testmodel = 'fg3')
#y ~ var1 + var2 + f(testm, testmodel = "test", testgraph = g, truetest1 = TRUE,
# truetest2 = TRUE) + f(testm2, testmodel = "fg") + f(testm3, testmodel = 'fg3')
#y ~ var1 + var4 + f(testm, testmodel = "test", testgraph = g, truetest1 = TRUE,
# truetest2 = TRUE) + f(testm2, testmodel = "fg") + f(testm3, testmodel = 'fg3')
#etc.....
If I simplify mm_exp, I can get something similar to what I want using reformulate (ignoring combn for now):
mm_exp_simplify <- expression(
f(testm, testmodel = 'test', testgraph = g),
f(testm2, testmodel = 'fg'),
f(testm3, testmodel = 'fg3')
)
reformulate(c(vars, sapply(mm_exp_simplify, deparse)), "y")
# y ~ var1 + var2 + var3 + var4 + f(testm, testmodel = "test",
# testgraph = g) + f(testm2, testmodel = "fg") + f(testm3,
# testmodel = "fg3")
but if I add back in truetest1 = TRUE, truetest2 = TRUE it causes issues:
mm_exp <- expression(
f(testm, testmodel = 'test', testgraph = g, truetest1 = TRUE, truetest2 = TRUE),
f(testm2, testmodel = 'fg'),
f(testm3, testmodel = 'fg3')
)
reformulate(c(vars, sapply(mm_exp, deparse)), "y")
# Error in reformulate(c(vars, sapply(mm_exp, deparse)), "y") :
# 'termlabels' must be a character vector of length at least one
I also tried using quote but had a similar problem:
mm_quote <- quote(
f(testm, testmodel = 'test', testgraph = g, truetest1 = TRUE, truetest2 = TRUE) +
f(testm2, testmodel = 'fg') + f(testm3, testmodel = 'fg3')
)
as.formula(paste0("y ~ ", paste(paste(vars, collapse = "+"), deparse(mm_quote), sep = "+")))
# Error in parse(text = x, keep.source = FALSE) :
# <text>:2:39: unexpected '='
# 1: y ~ var1+var2+var3+var4+f(testm, testmodel = "test", testgraph = g, truetest1 = TRUE,
# 2: y ~ var1+var2+var3+var4+ truetest2 =
# ^
Does anyone have suggestions on how to include the truetest1 = TRUE, truetest2 = TRUE and also how to get the combn versions of the formulas?
thanks
To solve the first problem, you need to use deparse1 instead of deparse. Like this:
reformulate(c(vars, sapply(mm_exp, deparse1)), "y")
#> y ~ var1 + var2 + var3 + var4 + f(testm, testmodel = "test",
#> testgraph = g, truetest1 = TRUE, truetest2 = TRUE) + f(testm2,
#> testmodel = "fg") + f(testm3, testmodel = "fg3")
About your second problem, first you can create all the combinations with all possible lengths, then you can create a list of all formulas in this way:
# all vars combinations
vars_comb <- lapply(seq_along(vars), function(n) combn(vars, n, simplify = FALSE))
vars_comb <- unlist(vars_comb, recursive = FALSE)
# all formulas
lapply(vars_comb, function(v) reformulate(c(v, sapply(mm_exp, deparse1)), "y"))
The reason behind it is related to the default value of the argument width.cutoff which is width.cutoff = 60L in deparse and width.cutoff = 500L in deparse1.
Just look at this:
# output with deparse
deparse(expression(f(testm, testmodel = 'test', testgraph = g, truetest1 = TRUE, truetest2 = TRUE)))
#> [1] "expression(f(testm, testmodel = \"test\", testgraph = g, truetest1 = TRUE, "
#> [2] " truetest2 = TRUE))"
# output with deparse and width.cutoff forced to 500
deparse(expression(f(testm, testmodel = 'test', testgraph = g, truetest1 = TRUE, truetest2 = TRUE)),
width.cutoff = 500)
#> [1] "expression(f(testm, testmodel = \"test\", testgraph = g, truetest1 = TRUE, truetest2 = TRUE))"
# output with deparse1
deparse1(expression(f(testm, testmodel = 'test', testgraph = g, truetest1 = TRUE, truetest2 = TRUE)))
#> [1] "expression(f(testm, testmodel = \"test\", testgraph = g, truetest1 = TRUE, truetest2 = TRUE))"
The first deparse creates a vector of length 2 which interferes with reformulate because it creates components of the formula that are not complaint.
If you have R 3.6 like you said in the comments, deparse1 is not available.
Therefore you need to set width.cutoff = 500L inside deparse.
The solution will look like this:
# first issue
reformulate(c(vars, sapply(mm_exp, deparse, width.cutoff = 500L)), "y")
# second issue
vars_comb <- lapply(seq_along(vars), function(n) combn(vars, n, simplify = FALSE))
vars_comb <- unlist(vars_comb, recursive = FALSE)
lapply(vars_comb, function(v) reformulate(c(v, sapply(mm_exp, deparse, width.cutoff = 500L)), "y"))