Changes to Celeus that you can see or use, dated to the day they went live. Follow along with the RSS feed.
, New
Suggested analyses now shows a Table 1 of your data as soon as you open it, and it does not count as an analysis run.
The “Describe your data” strip on Suggested analyses now builds Table 1 for you as soon as the page opens, from the de-identified version of your dataset and the columns you have in scope. Pick a grouping column and the table is rebuilt split by that group.
Building it does not use one of your analysis runs.
, Improved
Uploads show how much of the file has been sent, then a processing step, and you can cancel while the file is still sending.
The upload area now shows a progress bar while your file is sent, then a processing step while the dataset is read. You can cancel an upload until the file has finished sending.
The drop area also shows the size limit and the file types it accepts before you choose a file, and a failed upload shows an error code you can quote if you contact us.
, Improved
Columns you reviewed and kept now show as kept by you, and each decided column says whether it is sent to the AI or held back.
When the scanner flags a column as a possible identifier and you decide to keep it unchanged, the Profile table now marks it “Kept by researcher” and counts it as held back from the AI. Every column you have decided on also carries a second label saying whether it is sent to the AI or held back.
Attesting that a flagged column is not an identifier now asks for a short reason, so the decision is recorded with why it was made.
, New
Welch, Student and paired t-tests can now test whether two means are equivalent within a margin you set.
When you configure a Welch, Student or paired t-test you can now enter an equivalence margin. Celeus then runs two one-sided tests (Schuirmann, 1987) and reports whether the difference lies inside the margin, next to the usual test result, in the results and in the Methods paragraph.
A margin combined with a one-sided alternative is shown as a conflict before the run, because the two ask different questions.