teaching-phd-skills

Why this exists

Spreadsheets are where most lab data first lands and where most silent errors are born: a formula dragged one row short, a sort that moved one column and not its neighbours, a constant typed into a cell six months ago that nobody can now explain. Reviewers never see the spreadsheet — they see the number it produced. And AI is very good at generating Excel formulas, and equally good at generating ones that are subtly wrong: an absolute reference that should have been relative, a lookup returning the first match when you needed the exact one. The formula runs. It produces a number. Nothing announces the error.

External material

Microsoft’s Excel training — the XLOOKUP, PivotTable and named-range lessons. It covers the mechanics of each feature. It says nothing about how a spreadsheet should be organised when the output is going into a paper, which is the mechanic below.

Governing part

None directly. Once the spreadsheet is an input to a repo, the standard’s rule on immutable inputs (Part 1) governs the file.

The mechanic

One real dataset from your own work. Not a sample.

Structure it so every number is traceable.

1. Raw data on its own sheet, never edited. Not a correction, not a unit change.
2. Name the range.
3. Every transformation on a second sheet, referencing the first by name.
4. One pivot answering a question you actually have.
5. One chart with labelled axes and units.

The rule underneath: any cell a reader cannot trace back to the raw sheet is a number you cannot defend. A bare constant inside a formula is the usual way that happens — put it in a labelled cell and reference it.

Then get AI to write something you could not have written yourself — a nested INDEX/MATCH, a dynamic array, a small VBA macro. Before you trust it:

• Annotate every argument in your own words. Not the documentation’s. If you cannot say what an argument does, you do not have the formula, you have a guess that works.
• Break it on purpose. Delete a row, add a duplicate key, blank a cell. Write your predicted output before running it.

Getting the prediction wrong is more useful than getting it right, and you should say so rather than quietly fixing the note.

Deliverable

The .xlsx plus the annotation and the prediction, in that week’s WA entry.

Competency check

• Raw sheet exists and no formula writes to it
• At least one named range used in a formula
• No bare numeric constant inside a formula except 0, 1, and documented unit conversions
• One pivot answering a stated question; one chart with axes labelled and units given
• An AI-generated formula, with every argument annotated in the student’s own words
• A deliberate break, with the prediction shown before the result

Turning it in

This replaces your arithmetic entry for the week you do it. Submit it to that week’s WA assignment in Canvas — the same one you would have submitted anyway. It is not an extra piece of work on top of the log.

• In the text box: what you could not have written yourself, whether your prediction held, and one error the structure caught.
• You hand in: An .xlsx attached to the entry, plus a text-box write-up: raw sheet nothing writes to, transform sheet, one pivot, one chart with units, and your line-by-line annotation of the AI-written formula

You do not submit anything to the milestone yet. When all five skills in M1 are done, submit the M1 — The Workbench assignment, which is just an index saying which week each one went in.

This site is open source. Improve this page.