When a graphml file is imported to Server, and then exported from Server (as either csv or graphml), categorical data with numerical values do not export correctly. Categorical values which are strings export as expected.
To replicate:
create a protocol with a categorical variable. Give some of the options an integer value (e.g. 1, 0, 99).
Add data to a session, making sure to select some of the integer value options, so they are expected to be true on export.
Export as graphml.
Import the graphml file to Server.
Export the session from Server.
Categorical variables with integer values are all false.
I suspect this is a "2" is not 2 issue, in which somewhere in the pipeline of import/export in Server, graphml data looks like a string instead of an integer. Interestingly, the graphs within Server all show the categorical data as expected, so either graphs are compensating values or it is after import that numeric values are misinterpreted.
Note: this is only an issue for graphml imports. If a session is exported directly from Interviewer to Server, all data is exported correctly.
Workarounds include:
export from Interviewer directly to Server, skipping the graphml import
only use string values for categorical variables
export csv files in addition to the graphml from Interviewer
When a graphml file is imported to Server, and then exported from Server (as either csv or graphml), categorical data with numerical values do not export correctly. Categorical values which are strings export as expected.
To replicate:
1
,0
,99
).true
on export.false
.I suspect this is a "2" is not
2
issue, in which somewhere in the pipeline of import/export in Server, graphml data looks like a string instead of an integer. Interestingly, the graphs within Server all show the categorical data as expected, so either graphs are compensating values or it is after import that numeric values are misinterpreted.Note: this is only an issue for graphml imports. If a session is exported directly from Interviewer to Server, all data is exported correctly.
Workarounds include: