abbreviating words in a model formula

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf
Of William Dunlap
Sent: Monday, July 08, 2013 11:02 AM
To: Michael Friendly; R-help
Subject: Re: [R] abbreviating words in a model formula

Try using all.names() to get all the names in the formula.  E.g.,

f <- function (formula, minNameLength = 2, abbreviateFunctionNames = FALSE)
{
    names <- all.names(formula, functions = abbreviateFunctionNames)
    abbrNames <- lapply(abbreviate(names, minlength = minNameLength),
        as.name)
    deparse(do.call("substitute", list(formula, abbrNames)))
}

used as

  > f(MyResponse ~ log(FirstPredictor) + sqrt(SecondPredictor))

  [1] "MR ~ log(FP) + sqrt(SP)"

  > f(MyResponse ~ log(FirstPredictor) + sqrt(SecondPredictor), min=4)

  [1] "MyRs ~ log(FrsP) + sqrt(ScnP)"

  > f(MyResponse ~ log(FirstPredictor) + sqrt(SecondPredictor),

abbreviateFunctionNames=TRUE)
  [1] "MR ~ lg(FP) + sq(SP)"

You could put that in a loop that stopped when nchar(f(...)) got small enough.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf
Of Michael Friendly
Sent: Monday, July 08, 2013 10:36 AM
To: R-help
Subject: [R] abbreviating words in a model formula

For an application, I need to get a character string representation of
the formula or
model call for glm objects, but also, for labeling output and plots, I
want to be able
to abbreviate the words (variables) in model terms.  This requires some
formula
magic that I can't quite get, in particular extracting the terms from a
formula and
then the words in each term.

Perhaps there is some code for something similar
I haven't found yet, or someone can suggest how to do this.

A runnable example to show what I mean:

Freq <- c(68,42,42,30, 37,52,24,43,
     66,50,33,23, 47,55,23,47,
     63,53,29,27, 57,49,19,29)

Temperature <- gl(2, 2, 24, labels = c("Low", "High"))
Softness <- gl(3, 8, 24, labels = c("Hard","Medium","Soft"))
M.user <- gl(2, 4, 24, labels = c("N", "Y"))
Brand <- gl(2, 1, 24, labels = c("X", "M"))

detg <- data.frame(Freq,Temperature, Softness, M.user, Brand)
detg.m0 <- glm(Freq ~ M.user*Temperature*Softness +
Brand*M.user*Temperature,
        family = poisson, data = detg)

detg.m1 <- glm(Freq ~ (M.user + Temperature + Softness + Brand),
        family = poisson, data=detg)

detg.m2 <- glm(Freq ~ (M.user + Temperature + Softness + Brand)^2,
        family = poisson, data=detg)

detg.m2a <- update(detg.m1, . ~ .^2)

In plot.lm, I found the following code to extract the model call from a
glm object as
a string and abbreviate it to a total length <=75.  I need shorter total
length,
by abbreviating individual words in the model call, so the approach has to
at least extract the terms in the model and then abbreviate the words in
each term.

# from plot.lm: get model call as a string
# TODO: how to use abbreviate to abbreviate the words in the model terms???
mod.call <- function(x, max.len=75) {
         cal <- x$call
         if (!is.na(m.f <- match("formula", names(cal)))) {
             cal <- cal[c(1, m.f)]
             names(cal)[2L] <- ""
         }
         cc <- deparse(cal, max.len+5)
         nc <- nchar(cc[1L], "c")
         abbr <- length(cc) > 1 || nc > max.len
         cap <- if (abbr)
             paste(substr(cc[1L], 1L, min(max.len, nc)), "...")
         else cc[1L]
         cap
}

Tests, & WANTED, say with max length of each word in the string <= 6 &
maximum total
length <= 40

 > mod.call(detg.m0)

[1] "glm(Freq ~ M.user * Temperature * Softness + Brand * M.user *
Temperature)"

WANTED, somthing like:
"glm(Freq ~ M.user * Temp * Softne + Brand * M.user * Temp)"

 > mod.call(detg.m2a)

[1] "glm(Freq ~ M.user + Temperature + Softness + Brand +
M.user:Temperature + M ..."

 >
 > mod.call(detg.m2a, max.len=200)

[1] "glm(Freq ~ M.user + Temperature + Softness + Brand +
M.user:Temperature + M.user:Softness + M.user:Brand +
Temperature:Softness + Temperature:Brand + Softness:Brand)"

 >

WANTED, somthing closer to
"glm(Freq ~ M + Tmp + Sft + Brnd + M:Tmp + M.:Sft + M.us:Brnd + Tmp:Sft
+ Tmp:Brnd + Sft:Brnd)"

TIA
-Michael



--
Michael Friendly     Email: friendly AT yorku DOT ca
Professor, Psychology Dept. & Chair, Quantitative Methods
York University      Voice: 416 736-2100 x66249 Fax: 416 736-5814
4700 Keele Street    Web:   http://www.datavis.ca
Toronto, ONT  M3J 1P3 CANADA