Writing an R function for models in survey analysis-CodePudding

Data set up

library(survey)
library(gtsummary)
 
#Load data 
data(api)

#Declare survey design 
dclus1<-svydesign(id=~dnum, weights=~pw, data=apiclus1, fpc=~fpc)

I would like to construct a function that takes in an independent variable and produces a gt summary table with the results.

For example; I want to run a linear model to predict having "awards" based on having "both" parents in the household. I tried this:

mod_create <- function(y){
  model <- svyglm(y~both, design = dclus1, family = quasibinomial())
  tbl <- tbl_regression(model, exponentiate = TRUE) %>% add_global_p()
  return(tbl)
}

mod_create(awards)

I get this error:

 Error in svyglm.survey.design(y ~ awards, design = dclus1, family = quasibinomial()) : 
 all variables must be in design= argument
  4.stop("all variables must be in design= argument")
  3.svyglm.survey.design(y ~ awards, design = dclus1, family = quasibinomial())
  2.svyglm(y ~ awards, design = dclus1, family = quasibinomial())
  1.mod_create(both)

CodePudding user response：

You should be able to use reformulate here:

library(survey)
library(dplyr)
library(gtsummary)

data(api)
dclus1 <- svydesign(id =  ~ dnum, weights =  ~ pw, data = apiclus1, fpc =  ~ fpc)

mod_create <- function(y){
  model <- svyglm(reformulate(paste0(y, " ~ both")), 
                  design = dclus1, family = quasibinomial())
  tbl <- tbl_regression(model, exponentiate = TRUE) %>% add_global_p()
  return(tbl)
}

mod_create("awards") # note the ""