Created attachment 1487 [details]
.csv input data file to recreate issue
Calling fix() on a data frame seems to be rounding one of the columns (response_a) of the data file attached. We have only seen this issue in these data. It does cause important changes in later significance testing. The issue can be recreated by:
> data <- read.csv("fix_issue_data.csv")
> write.csv(data, file = "post_no_data_fix")
> write.csv(data, file = "post_data_fix")
I can confirm the problem: values are displayed with rounding, then changed even if the user never edits anything.
A workaround is to set the "digits" option to a large value, e.g. options(digits=15), before calling fix() or edit() or data.edit() on the dataframe.
This only happens in the R.app data editor, not in X11 or Windows editors.