Response Quality Indicators for Survey Research.

resquin

About

resquin (response quality indicators) provides functions to calculate survey data quality indicators to help identifying low-quality responses (Bhaktha, Silber, and Lechner 2024; Curran 2016; Vaerenbergh and Thomas 2013). resp_styles(), resp_distributions(), resp_nondifferentiation() and resp_patterns() provide response quality indicators geared towards multi-item scales or matrix questions. Both multi-item scales and matrix questions present survey respondents with multiple questions which have the same response format, meaning the same number and labeling of response options.

At the moment, resquin provides four functions:

• resp_styles() - Calculates response style indicators (e.g. extreme response style or middle response style).
• resp_distributions() - Calculates response distribution indicators (e.g. intra-individual mean and standard deviation over a set of survey questions).
• resp_nondifferentiation() - Calculates response nondifferentiation indicators. Nondifferentiation indicators primarily measure straightlining. The indicators differ in how straightlining is operationalized.
• resp_patterns() - Calculates response pattern indicators (e.g. long string analysis).

For information on how to use resquin see the vignettes Getting started with resquin and resquin in practice.

resquin is still under active development. Please use github issues to file questions and bug reports or send them directly to [email protected]. We are happy to receive feedback!

Installation

resquin is available via CRAN and github. To install resquin from CRAN or github, you can use one of the following commands:

# Install resquin via CRAN
install.packages("resquin")

# Install development version of resquin with devtools
devtools::install_github("https://github.com/MatRoth/resquin")

# Install development version of resquin with pak
pak::pak("https://github.com/MatRoth/resquin")

Getting started

To use resquin, supply a data frame containing survey responses in wide format to either resp_styles(), resp_distributions(), resp_nondifferentiation(), resp_patterns().

# load resquin
library(resquin)

# A test data set with three items and ten respondents
testdata <- data.frame(
  var_a = c(1,4,3,5,3,2,3,1,3,NA),
  var_b = c(2,5,2,3,4,1,NA,2,NA,NA),
  var_c = c(1,2,3,NA,3,4,4,5,NA,NA))

testdata
#>    var_a var_b var_c
#> 1      1     2     1
#> 2      4     5     2
#> 3      3     2     3
#> 4      5     3    NA
#> 5      3     4     3
#> 6      2     1     4
#> 7      3    NA     4
#> 8      1     2     5
#> 9      3    NA    NA
#> 10    NA    NA    NA

# Calculate response style indicators per respondent
resp_styles(x = testdata,
            scale_min = 1,
            scale_max = 5) |> # Specify scale minimum and maximum
  round(2)
#> # Number of missings due to min_valid_responses equal to 1: 4
#> # A data frame:                                             10 × 6
#>       id   MRS   ARS   DRS   ERS  NERS
#>    <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#>  1     1  0     0     1     0.67  0.33
#>  2     2  0     0.67  0.33  0.33  0.67
#>  3     3  0.67  0     0.33  0     1   
#>  4     4 NA    NA    NA    NA    NA   
#>  5     5  0.67  0.33  0     0     1   
#>  6     6  0     0.33  0.67  0.33  0.67
#>  7     7 NA    NA    NA    NA    NA   
#>  8     8  0     0.33  0.67  0.67  0.33
#>  9     9 NA    NA    NA    NA    NA   
#> 10    10 NA    NA    NA    NA    NA

# Calculate response distribution indicators per respondent
resp_distributions(x = testdata) |>
  round(2)
#> # Number of missings due to min_valid_responses equal to 1: 4
#> # A data frame:                                             10 × 7
#>       id  n_na prop_na ii_mean ii_sd ii_median mahal
#>    <dbl> <dbl>   <dbl>   <dbl> <dbl>     <dbl> <dbl>
#>  1     1     0    0       1.33  0.58         1  2.04
#>  2     2     0    0       3.67  1.53         4  1.6 
#>  3     3     0    0       2.67  0.58         3  1.38
#>  4     4     1    0.33   NA    NA           NA NA   
#>  5     5     0    0       3.33  0.58         3  0.97
#>  6     6     0    0       2.33  1.53         2  1.38
#>  7     7     1    0.33   NA    NA           NA NA   
#>  8     8     0    0       2.67  2.08         2  1.88
#>  9     9     2    0.67   NA    NA           NA NA   
#> 10    10     3    1      NA    NA           NA NA

# Calculate response nondifferentiation indicator per respondent
resp_nondifferentiation(x = testdata) |> 
  round(2)
#> # Number of missings due to min_valid_responses equal to 1: 4
#> # A data frame:                                             10 × 5
#>       id simple_nondifferentiation mean_root_pairs max_identical_rating
#>    <dbl>                     <dbl>           <dbl>                <dbl>
#>  1     1                         0            1                    0.67
#>  2     2                         0            0.21                 0.33
#>  3     3                         0            1                    0.67
#>  4     4                        NA           NA                   NA   
#>  5     5                         0            1                    0.67
#>  6     6                         0            0.21                 0.33
#>  7     7                        NA           NA                   NA   
#>  8     8                         0            0                    0.33
#>  9     9                        NA           NA                   NA   
#> 10    10                        NA           NA                   NA   
#> # ℹ 1 more variable: scale_point_variation <dbl>

# Calculate response pattern indicators
resp_patterns(x = testdata) |> 
  round(2)
#> # Number of missings due to min_valid_responses equal to 1: 4
#> # A data frame:                                             10 × 4
#>       id n_transitions mean_string_length longest_string_length
#>    <dbl>         <dbl>              <dbl>                 <dbl>
#>  1     1             2                  1                     1
#>  2     2             2                  1                     1
#>  3     3             2                  1                     1
#>  4     4            NA                 NA                    NA
#>  5     5             2                  1                     1
#>  6     6             2                  1                     1
#>  7     7            NA                 NA                    NA
#>  8     8             2                  1                     1
#>  9     9            NA                 NA                    NA
#> 10    10            NA                 NA                    NA

For a more information on how to use resquin see the vignettes Getting started with resquin and resquin in practice.

Researchers have developed a large number of response quality indicators. resquin only covers a few of them. The following list contains packages which provide alternative implementations of the indicators used in resquin or provide other indicators not used in resquin:

• Implementation of (Kim et al. 2019) measures: straightliner

• Implementation of indices from the careless responding literature: careless

• Autocorrelation and pattern indices: responsePatterns

Below are packages which use IRT to model and estimate response styles and individual response quality:

• PCMRS

• FitResponseStyles

• Model Multivariate Ordinal Responses Including Response Styles

• Person fit indices (Perfit)

Finally, there are packages for pattern checking:

• pattern.checks

• bpa

References