josephholler
commented
2 years ago Another option may be to use the SPSS file format. Although it is proprietary, it contains both values and labels and the GNU PSPP open source project has documented the file format.
The haven package can import this, with read_sav, resulting in a data frame having numerical values and qualitative labels. Similar, but not quite the same as using vectors. The labelled and expss packages help work with this type of data structure. IDK if you think it's easier to use the previous style of code to reorder factors from our CSV data import (advantage: all the data cleaning code will still work), or if its easier to download the data in SPSS format and use functions from haven and expss and/or labelled to manipulate the data.
These expss examples make it look very easy to do some good-looking data exploration in R: https://cran.r-project.org/web/packages/expss/vignettes/tables-with-labels.html
These labelled functions make it look easy to convert to data frames with factors or do other useful things:
e.g. df %>% to_factor() is to convert all labelled vectors and only labelled vectors into factors: https://raw.githubusercontent.com/rstudio/cheatsheets/main/labelled.pdf
Is it possible to preserve a more sensible order for the factors of ordinal variables in our data frame?
If factors are never ordered correctly through the workflow of exporting from Qualtrics and into R, then we can manually override the order for a single variable with code like this:
survey_resp$Q8_1 <- factor(survey_resp$Q8_1, c("Strongly disagree", "Disagree", "Agree", "Strongly agree", ""), ordered = TRUE)and then save the data also as an
rdsfile in addition to CSV. Hopefully that approach can be scaled up easily with loops and lists for our different types of ordinal responses.