Nine o'clock, campus office, the hallway lights already on their timer. Two windows are open. One holds the manuscript, drawn from a dissertation study finished eighteen months ago. The other holds the reviewers' comments. Reviewer 2 wants reliability coefficients for the 24-item climate scale, subscale by subscale, not one overall number in a footnote. From Reviewer 1: correct Table 2, because the standard deviations in the text and the standard deviations in the table disagree. Friday is the deadline.
The statistics here are not difficult. An applied researcher can describe the whole job in one breath: run the reliability analysis, look at the item level, fix the table, write the response letter. Ten minutes of thinking, maybe twenty. The evening will not be twenty minutes.
What the Night Actually Costs
Most of the cost has nothing to do with statistics. It goes to reopening a project that has been closed since the defense, finding which of the four files was the one that produced the published numbers, and waiting on an update that decided tonight was the night. After that comes the part nobody counts: reading numbers off one screen and typing them into a Word table, then checking the typing, then checking it again, because a transposed digit in Table 2 is exactly the error that got flagged the first time.
Anyone who has done a revise and resubmit knows the shape of this evening. The intellectual work is finished by 9:20. Everything after that is transfer, formatting, and verification, and it runs until midnight.
The Machine Should Be Quiet About Itself
A good tool at 9 pm asks for nothing. No environment to assemble, no packages to reconcile, no dialog explaining what has changed since the last session. Open the app, open the file, work.
Apple Silicon settled the performance side of this. A reliability analysis on 214 cases and 24 items finishes before you have finished reading the reviewer's sentence, and it finishes on the same machine holding the manuscript, the comments, and the response letter. Nothing has to reach a server. On a laptop in a quiet building, that also means the battery is the only resource anyone is watching.
Alpha, Item 17, and a Table You Do Not Retype
This is where Quanta does its part. Load the CSV. Choose the reliability analysis and pick the subscales. What comes back is not a single coefficient: alpha for the full scale and for each subscale, the item-total correlation for every item, and the alpha you would have if that item were removed.
Item 17 shows an item-total correlation of .11 while the rest of its subscale sits between .42 and .63. Removing it moves that subscale from .68 to .79. Now Reviewer 2's question has an answer with a number attached, and the answer can go in the response letter tonight.
Two caveats belong in the same paragraph as that number. A weak item-total correlation is evidence, not a verdict; item 17 may be measuring something the theory says belongs in the scale, and dropping it is a decision you defend in the text. Alpha reports internal consistency, which is not validity. Reviewers who read carefully will ask about both, and they should.
Table 2 comes out formatted to APA 7 and lands in Word as an editable table, not a picture of one. The standard deviations in the table are the standard deviations from the analysis, so the disagreement Reviewer 1 caught cannot survive the export. That single fact removes the retyping, and with it the class of error that produced the correction request.
Quanta does not think for you. It removes the translation between your question and your evidence, and that is the whole of what it removes. Which items belong in the scale, whether the revision answers the critique, how much to concede in the letter: those stay with the researcher.
What Friday Looks Like
The realistic version of this evening ends around 10:30 with a response letter drafted point by point, each reviewer comment answered with a number, a table, or a sentence explaining why the change was not made. Friday becomes a proofreading day instead of a rescue.
None of that makes the reviewers easier to satisfy. The scale still has a weak item, and someone still has to argue for the design. What changes is where the hours went. Three of them went to the argument rather than to retyping standard deviations at eleven at night, and the argument is the only part of the evening the field will ever see.
Quanta runs on the Mac, offline, and is written in Swift for Apple Silicon and checked against NIST reference datasets. Available at relicheck.com and on the Mac App Store, with a free 30-day trial.