R:Pass Argument to Glm Inside an R Function

R : Pass argument to glm inside an R function

Solution with substitute (@DWin suggestion).

function(y, x, dataset, weights){
  f <- substitute(glm(y~x, data=dataset, weights=weights, family=binomial))
  logLik(eval(f))
}

How to pass function arguments to R model object

Try this:

grid = (cbind(c('wl', 'livingd', 'deceasedt'), c('wl_time', 'ld_time', 'dec_time')))

for (k in 1:nrow(grid)){

  f=function(y=as.name(grid[k,1]), offset=as.name(grid[k,2])){
    m=glm(get(y)~chain_class2+sex_new+age_cat+race_new, 
                      family=poisson(link='log'), 
                      data=poissonset, 
                      offset=log(get(offset)/12))
  }
}

R - how to pass formula to a with(df, glm(y ~ x)) construction inside a function

Try wrapping the formula in as.formula

fit_model_mi = function(formula) {
    with(mtcars.imp, glm(as.formula(formula)) )
}

Seems to work:

> fit_model_mi("mpg ~ cyl")
call :
with.mids(data = mtcars.imp, expr = glm(as.formula(formula)))

call1 :
mice(data = mtcars, m = 5)

nmis :
 mpg  cyl disp   hp drat   wt qsec   vs   am gear carb 
   0    0    0    0    1    0    0    0    0    0    0 

analyses :
[[1]]

Call:  glm(formula = as.formula(formula))

Coefficients:
(Intercept)          cyl  
     37.885       -2.876  

Degrees of Freedom: 31 Total (i.e. Null);  30 Residual
Null Deviance:      1126 
Residual Deviance: 308.3    AIC: 169.3

pass family= to step() via glm() programmatically

I think the following solution, based on eval, bquote and .() might solve your problem.

I also have R-version 3.2.4 installed, and I got the exact same error you got from your code. The solution below made it work at my computer.

getCoef <- function(formula, 
                family = c("gaussian", "binomial"),
                data){

    model_fam <- match.arg(family, c("gaussian", "binomial"))

    fit_null <- eval(bquote(
        glm(update(.(formula),".~1"), 
            family = .(model_fam), 
            data = .(data))))

    message("So far so good")

    fit_stepBIC <- step(fit_null, 
                        formula, 
                        direction="forward",
                        k = log(nrow(data)),
                        trace=0)

    message("Doesn't make it this far")

    fit_stepBIC$coefficients
}

# returns error 'model_fam' not found 
 getCoef(formula = Petal.Length ~ Petal.Width + Species,
        family = "gaussian",
        data = iris)

So far so good
Doesn't make it this far
      (Intercept) Speciesversicolor  Speciesvirginica       Petal.Width 
         1.211397          1.697791          2.276693          1.018712   

Passing Argument to lm in R within Function

Building on the comments, gear isn't defined globally. It works inside the stand-alone lm call as you specify the data you are using, so lm knows to take gear from df.

Howver, gear itself doesn't exist outside that stand-alone lm function. This is shown by the output of gear

> gear
Error: object 'gear' not found

You can pass the gear into the function using df$gear

weightvar <- df$gear
ols <- olswrapper(mpg ~ cyl + qsec, weightvar , df = df)

Passing weights to glm() using rlang

The issue is that glm() can recognize an expression being provided to its weights argument, but doesn't support quasiquotation, because it uses the base quote() / substitute() / eval() mechanisms instead of rlang. This causes problems for nested expression arithmetic.

One way to get around it is to compose the entire glm expression, then evaluate it. You can use ... to supply optional arguments.

myglm2 <- function( .data, y, x, weights, ... ) {
  myglm <- expr( glm(!!enexpr(y) ~ !!enexpr(x), data=.data, 
                      weights = !!enexpr(weights), ...) )
  eval(myglm)
}

myglm2(mydata, outcome, group)
# Call:  glm(formula = outcome ~ group, data = .data)

myglm2(mydata, outcome, group, wgts)
# Call:  glm(formula = outcome ~ group, data = .data, weights = wgts)

myglm2(mydata, outcome, group, wgts, subset=7:10)
# Call:  glm(formula = outcome ~ group, data = .data, weights = wgts, 
#     subset = ..1)
# While masked as ..1, the 7:10 is nevertheless correctly passed to glm()

To follow @lionel's suggestion, you can encapsulate the expression composition / evaluation into a standalone function:

value <- function( e ) {eval(enexpr(e), caller_env())}

myglm2 <- function( .data, y, x, weights, ... ) {
  value( glm(!!enexpr(y) ~ !!enexpr(x), data=.data, 
              weights = !!enexpr(weights), ...) )
}

How to make glm object within a function take input variable names and not parameter names?

You can use the as.formula() function to transform a string with your formula before calling glm(). This will solve your question (How do I make the glm objects refer to actual variables), but I'm not sure if it is enough for the calling cv.glm later (I couldn't reproduce your code here, without errors). To be clear, you replace the line

mod = glm(dep~poly(indep,i),data=dat)

with something like:

myexp = paste0(dep, "~ poly(", indep, ",", i, ")")

mod = glm(as.formula(myexp), data=dat)

it's required then to make the variables dep and indep to be characters with names of the variables that you want to refer to (e.g. indep="displ").

How to pass bunch of elements as an argument in R?

The reformulate() function is what you need!

mylabels = c("Lag1","Lag2","Lag3","Lag4")
myresponse = "Direction"
reformulate(mylabels, myresponse)

Direction ~ Lag1 + Lag2 + Lag3 + Lag4

---

Related Topics