In DataS-DHSC/consultations: Re-usable functions for describing the responses to a public consultation or call for evidence

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(consultations)
data(dummy_response, package = "consultations")

This package comes with some generated dummy data. This contains a few categorical columns, one column representing a multi-choice tickbox question, and one column with randomly generated free text.

knitr::kable(dummy_response[1:5,])

Categorical analysis

Any questions that are multiple choice, and/or don't have too many free-text responses, can be described and visualised in frequency tables, bar plots, and short text. This package provides functions to do this quickly.

First, the question_types() function can detect from some rules of thumb which columns are which question type. This function returns a list with three elements: categorical, multichoice, and freetext. Each element contains a vector of column names, containing the column names which fall into each group.

survey_question_types(dummy_response)

Second, based on these question types we can calculate appropriate frequency tables. For this, we can use the response_tables() function. This returns a list, with an element for each column (ie. an element for each question). This function, by default, aggregates any answers that have fewer than 5 responses to avoid statistical disclosure: you can set this yourself using the min_n parameter.

survey_response_tables(dummy_response, survey_question_types(dummy_response), min_n = 5)

Third, we can describe these frequency tables as a bar plot. The survey_response_barplot() function generates a list of ggplot objects, one for each question. This is indiscriminate about question type, so for free text questions it will simply return a "Free text" bar.

This function can take the data directly, since it runs response_tables() under the hood.

survey_response_barplot(dummy_response)

Finally, we can describe this same data using text bullet points: how many possible responses were there, what was the most common, and what was the least common. The summary_text() function takes the output from response_tables, but only one element at a time, so you have to run it as part of an lapply call.

lapply(survey_response_tables(dummy_response, qtypes = survey_question_types(dummy_response)), survey_summary_text)

Free text analysis

If you have columns with free text, the quantitative options above will not return useful results. Instead, you can use the text_ set of functions provided in this package.

DataS-DHSC/consultations documentation built on Jan. 28, 2022, 1:56 a.m.