Working with variables in a function in R

Question

I am trying to create a function to run chi squared where I have to group by several groups. However, while the method works when it's not a function, I am having trouble turning into a function. As I'll be repeating the procedure multiple times, its seems worth doing, but I just can't get the function to recognise the "z" variable and always get the "Unknown or uninitialised column" warning.

Example is below.

library(tidyverse)
library(datasets)

#data
data(iris)
df<-iris%>%
  gather(Type, value, -Species)%>%
  separate(Type, c("type", "attribute"), sep="[.]")

#functions------------
frequency<-function(data, x, y, z){
  x <- enquo(x)
  y <- enquo(y)
  z <- enquo(z)

  data%>%
    filter(!is.na(!!x), !is.na(!!y), !is.na(!!z))%>%
    count(!!x, !!y, !!z)
}

group_chi<-function(data, x, y, z){
  x <- enquo(x)
  y <- enquo(y)

  data %>%
    group_by(!! x) %>%
    nest() %>%
    mutate(M = map(data, function(dat){
      dat2 <- dat %>% spread(!! y, n)
      M <- as.matrix(dat2[, -1])
      row.names(M) <- dat2$'z' #I've done it like this becasue z <- enquo(z) and dat2$!!z doesn't work. jsut having it a z doesnt work either
      return(M)
    }))%>%
    mutate(pvalue = map_dbl(M, ~chisq.test(.x)$p.value)) %>%
    select(-data, -M) %>%
    ungroup()
}

#aplying them--------------------

test<-frequency(df, type, Species, attribute)
chi_test<-group_chi(test,  type, Species, attribute)#brings up warning
#> Warning: Unknown or uninitialised column: 'z'.

#> Warning: Unknown or uninitialised column: 'z'.

#test without the function=no warning. 
No_function<-test %>%
  group_by(type) %>%
  nest() %>%
  mutate(M = map(data, function(dat){
    dat2 <- dat %>% spread(Species, n)
    M <- as.matrix(dat2[, -1])
    row.names(M) <- dat2$attribute
    return(M)
  }))%>%
  mutate(pvalue = map_dbl(M, ~chisq.test(.x)$p.value)) %>%
  select(-data, -M) %>%
  ungroup()


# in the example the results are the same but....the warning message is of concern and the function doesn't output the same in a more compelx dataset.

chi_test 
#> # A tibble: 2 x 2
#>   type  pvalue
#>   <chr>  <dbl>
#> 1 Petal      1
#> 2 Sepal      1
No_function 
#> # A tibble: 2 x 2
#>   type  pvalue
#>   <chr>  <dbl>
#> 1 Petal      1
#> 2 Sepal      1
# what am I doing wrong?

^{Created on 2020-01-27 by the reprex package (v0.3.0)}

What am I doing wrong here?

r2evans · Accepted Answer · 2020-01-27 17:23:17Z

You can't use $ for an indirect column reference (as in dat2$'z'), instead use dat2[[z]]. When I replace that, there are no warnings/errors.

Try this version of your function instead:

group_chi<-function(data, x, y, z){
  x <- enquo(x)
  y <- enquo(y)

  data %>%
    group_by(!! x) %>%
    nest() %>%
    mutate(M = map(data, function(dat){
      dat2 <- dat %>% spread(!! y, n)
      M <- as.matrix(dat2[, -1])
      row.names(M) <- dat2[[z]]
      return(M)
    }))%>%
    mutate(pvalue = map_dbl(M, ~chisq.test(.x)$p.value)) %>%
    select(-data, -M) %>%
    ungroup()
}

and then call with the string:

chi_test <- group_chi(test,  type, Species, "attribute")

Alternatively, you can first z <- enquo(z) then pull(dat2, !!z) (as in @akrun's answer).

group_chi<-function(data, x, y, z){
  x <- enquo(x)
  y <- enquo(y)
  z <- enquo(z)

  data %>%
    group_by(!! x) %>%
    nest() %>%
    mutate(M = map(data, function(dat){
      dat2 <- dat %>% spread(!! y, n)
      M <- as.matrix(dat2[, -1])
      row.names(M) <- pull(dat2, !!z)
      return(M)
    }))%>%
    mutate(pvalue = map_dbl(M, ~chisq.test(.x)$p.value)) %>%
    select(-data, -M) %>%
    ungroup()
}
group_chi(test,  type, Species, attribute)
# # A tibble: 2 x 2
#   type  pvalue
#   <chr>  <dbl>
# 1 Petal      1
# 2 Sepal      1

akrun · Accepted Answer · 2020-01-27 17:26:02Z

0

We could also use z <- enquo(z), then make use of the select and pull to extract the column as a vector

group_chi<-function(data, x, y, z){
  x <- enquo(x)
  y <- enquo(y)
  z <- enquo(z)

  data %>%
    group_by(!! x) %>%
    nest() %>%
    mutate(M = map(data, function(dat){
      dat2 <- dat %>% spread(!! y, n)
      M <- as.matrix(dat2[, -1])

      row.names(M) <- dat2 %>% 
                           select(!!z) %>%
                           pull(1)
      return(M)
    }))%>%
    mutate(pvalue = map_dbl(M, ~chisq.test(.x)$p.value)) %>%
    select(-data, -M) %>%
    ungroup()
}

-checking

chi_test <- group_chi(test,  type, Species, attribute)
chi_test
# A tibble: 2 x 2
#  type  pvalue
#  <chr>  <dbl>
#1 Petal      1
#2 Sepal      1

With the newer versions of tidyverse, the curly-curly operator ({{}}) can replace the !!/enquo

group_chi<-function(data, x, y, z){

  data %>%
    group_by({{x}}) %>%
    nest() %>%
    mutate(M = map(data, function(dat){
      dat2 <- dat %>% spread({{y}}, n)
      M <- as.matrix(dat2[, -1])

      row.names(M) <- dat2 %>% 
                           pull({{z}})
      return(M)
    }))%>%
    mutate(pvalue = map_dbl(M, ~chisq.test(.x)$p.value)) %>%
    select(-data, -M) %>%
    ungroup()
}

chi_test <- group_chi(test,  type, Species, attribute)

edited Jan 27, 2020 at 17:26

answered Jan 27, 2020 at 17:19

akrun

891k38 gold badges590 silver badges700 bronze badges

1 Comment

ArturR2 Over a year ago

Thanks! You two are brilliant. It works perfectly now. I decided to use the curly-curly method with pull to cut down on code and so I would NOT have to call the string as in r2evans answer.

Collectives™ on Stack Overflow

Working with variables in a function in R

2 Answers 2

Comments

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related