In the following exercises, we will explore some of the options for dealing with missing values. As before, we will focus on functions from the tidyverse (more specifically, from dplyr plus one tidyr).

If you have not yet done so in your current R session, please load the same packages that we have used in the previous set of exercises (sjlabelled, tidyverse, haven; ideally also in that order).

We should also import the data again.

1

Import the ALLBUS 2021 .sav and assign it to an object named allbus_2021_missings. However, unlike in the previous set of exercises, for our missing value operations, we do not want to user-defined missing values from the SPSS file to be converted to NA when the data are imported. Same as before, we want to remove all labels.

To override the default behavior of converting user-defined missing values into NA, we need to provide an appropriate value for the user_na argument of the import function we need. To check what the appropriate value is, you can look at the function help file (?read_sav()).

solution 

allbus_2021_missings <- read_sav("./data/allbus_2021/ZA5280_v1-0-0.sav",
                        user_na = TRUE) %>% 
  remove_all_labels()

2

As a first exercise, let’s turn all negative values for the variable measuring satisfaction with democracy in Germany into NAs using a function from the dplyr package.

To to this, we need to combine mutate() with the dplyr function for recoding specific values as NA. The name of the variable is ps03 and the values we need to turn into NAs are: -42, -11, -9, -8.

solution 

allbus_2021_missings <- allbus_2021_missings %>% 
  mutate(ps03 = na_if(ps03, -42)) %>% 
  mutate(ps03 = na_if(ps03, -11)) %>% 
  mutate(ps03 = na_if(ps03, -9)) %>%
  mutate(ps03 = na_if(ps03, -8))

3

After recoding a set of values as NA for one variable, let’s now do the same for a whole dataframe. For this purpose, first create a new dataframe named contact that only contains variables assessing in which contexts respondents have contact with immigrants. Then convert the following values to NA for all variables in this dataframe: -42, -11, -10, -9.

The names of the variables we are looking for are: mc01, mc02, mc03, and mc04. For converting values to NA for all columns/variables in a dataframe, we do not need the mutate() function.

solution 

contact <- allbus_2021_missings %>% 
  select(mc01:mc04)

contact <- contact %>% 
  na_if(-42) %>% 
  na_if(-11) %>% 
  na_if(-10) %>% 
  na_if(-9)

4

As na_if() only takes only takes single values as its second argument (i.e., the value to replace with NA), let’s use a function from the sjlabelled function to achieve the same thing as in the previous task with fewer lines of code.

The function we are looking for can also be included in a pipe chain and takes a vector of values to be recoded as NA as its second (required) argument.

solution 

library(sjlabelled)

contact <- contact %>% 
  set_na(na = c(-42, -11, -10, -9))

5

As we final exercise, let’s build on what we did in task 2 and use a function from the tidyr package to answer the following question: How many of the respondents do not have a missing value for the item assessing satisfaction with democracy in Germany? This time, do not assign the result of your code to a new object.

To count the number of cases, you can use the base R function nrow() at the end of your pipe.

solution 

allbus_2021_missings %>% 
  drop_na(ps03) %>% 
  nrow()
## [1] 3523