Improved Gorilla data reporting format urgently needed!
M
Michael R Cunningham
All known statistical analysis programs, such as SPSS, SAS, R, and Systat expect the data from each subject to be represented by just one row of data, with each variable in a discrete column. Unfortunately, Gorilla's data output does not conform to standard analytic expectations in two ways:
1) Gorilla task files are produced only in "long form", in which data from a single subject are output in columns, rather than rows. This requires tedious data-wrangling procedures to reformat the data, which increases the chances of confusion and errors. Investigators earnestly plea that Gorilla reformat its output format to produce files that conform to the single row, multiple column convention.
2) Gorilla makes artificial distinctions between task data and questionnaire data and produces separate files for each. This requires that the multiple files be merged using a key variable to identify individual subjects. But this process is difficult because only questionnaire data files come in the short form, with a single row per subject. This again requires tedious data-wrangling procedures to reformat the task data, and then merge them with the questionnaire data, which increases the chances of confusion and errors. Investigators earnestly plea that Gorilla reformat its output format to produce files that conform to the single row per subject, multiple column convention, integrating data from across modes and nodes.
In addition:
3) It also would be helpful if Gorilla would allow investigators to set up data filters to exclude variables that the investigator does not wish to see or analyze. Reducing the number of variables in a data file is essential for analytic speed and investigator mental health.
4) Finally, it would be helpful if Gorilla would allow investigators to rename variable to something meaningful, so we can see at a glance if subjects are responding appropriately, or not. With Gorilla's name of variables with nonsense strings, quick and intuitive data monitoring is impossible.
Jade Pickering (Head of Product)
Hi Michael, thank you for taking the time to leave us your comments.
As with most experiment building software, data output is not designed to be ready for statistical analysis straight away. There will always be a degree of pre-processing - as with all studies - that is unique to the researcher's needs. For example, summarising data at the participant level and using your desired measure of central tendency, outlier detection and removal processes etc., before you reach the stage that your data is in the format of one row per participant.
We provide lots of guidance on how to do that on our support documentation. Many researchers make use of pivot tables in Excel but if you enjoy using R we also have guides on how to begin pre-processing your data with common R packages as well. Dr Emma James, a Gorilla user, has written a wonderful step-by-step guide that we highly recommend.
We're always looking to improve Gorilla, including data output, and will of course take your feedback on board.
Suzanna Forwood
Gorilla is not a data processing or pre-processing platform. All the functionality you ask for is provided by a data processing and pre-processing products. I use Excel, or R, or Jamovi depending on project.
R
Richard Kirkden
I use Excel pivot tables to process long form data.