manybabies / validator Goto Github PK

Documenting a workaround I've had to do to get the validator working on my system:

The 'studies' object doesn't seem to be in scope (?) when I run the server.R app. Copy-pasting lines 7-9 into my console to load that variable and then running again serves as a temporary solution.

(I originally thought this was a problem loading tidyverse, but it wasn't)

The validator currently is passing through a file (participant and trial files from Oslo) that has a bunch of extra columns in it. Are we checking for "extra" columns? I think we should be excluding (or at least putting up a warning) if a file has extra columns other than the ones specified.

should generate a random sample with appropriate format

In MB1 we asked people to use various not-valid data codes (not included, not collected, NA), but these are all collapsed in the analysis scripts to NA.

For MB3 specifically, and the validator in general - check how the "NA allowed = TRUE" validation works, and document it!

European (and other) locations use "," as the decimal marker. This may break the CSV based validator which was made by North Americans! Test this by having @cmfo500 create local Excel docs, exporting CSV, and testing in the shiny app.

To do: If it doesn't work, diagnose/create new issues to fix this!

Documentation for the condition variable was...inconsistent! To deal with this, allow FOUR values for the factor variable trial_type

same
different
consistent
inconsistent

(The analysis team will need to resolve this, but we want to avoid throwing a confusing error at contributors)

Lab IDs are really important for our datasets! Especially tracking & merging across the datasets. We've added reminders to the manual to check that, but also add a reference to the mega list here!

https://docs.google.com/spreadsheets/d/1O_c4QoRwBUQvpCABFC2XIlx4upsvHZm02VDV8xDUg8s/edit?usp=sharing

what should the desired functionality be? should we kick back errors or warnings for extra fields in the CSV?

Validator is throwing an error for 'medical_issue' and 'developmental_concern' for putting NA, which is the specification.

For future compatibility with Psych-DS, create JSON-LD Dataset compliant versions of the specifications, and then implement validator to be able to read them.

should render the specification so that it can be examined

If data in a particular column should meet some limitation (either type, e.g. must be numerical; or value, e.g. must be in {male, female, NA}), can the validator test for these conditions?

On testing the validator, it wasn't clear whether lowercasing the lab IDs was (a) intended to be enforced and (b) actually enforced.

Close when: MB3 analysis makes a call, someone tested the validator to see whether decision is implemented