What Good Statistical Coaching for Graduate Students Actually Looks Like

2 min read ▶ Listen to this post

"Statistics help for grad students" quietly covers two very different services: doing the analysis for a student, and coaching a student through doing it themselves. Both are legitimate depending on the situation — a student two days from a deadline with a broken model needs the first one. But a lot of what gets called "help" is the first one when the student actually needed the second, and the difference doesn't show up until the dissertation defense, when a committee asks the student to explain and defend a choice a consultant made for them.

What real coaching looks like

• Starting from the student's own question and data, not a generic textbook example — so the method makes sense in the exact context the student will have to defend it in later.
• Explaining the why, not just the how. Why this test, why this model — not just which function to call — so the student can make the next similar decision alone.
• Building habits that scale to a whole dissertation, not just solving today's problem: a script-based, reproducible workflow, a written analysis plan decided before the results come in, version control. A dissertation is a multi-year project, not a single assignment, and the habits that get built on chapter one are the ones that are still there for chapter four.
• Preparing the student to defend the choices — anticipating the "why didn't you just use X instead" question before the defense does, not after.
• Knowing when to step back and let the student work through something versus when being stuck has stopped being productive and started costing real time against a real deadline. Coaching isn't refusing to ever just answer the question.

Why it's worth asking for specifically

A student coached through their first multilevel model can run their second one alone. A student who was handed a finished analysis for the first one is in exactly the same position for the second — which is fine if this is a one-off project, and a real problem if it's chapter one of five.

If you're a student — or advising one — looking for the kind of help that's still useful after the current deadline passes, this is what our methods coaching is built around.

Related reading

• When Should You Involve a Statistician? (Earlier Than You Think) — the same question from the other direction: when to bring in outside help at all, and why earlier beats later.

One email whenever we publish something new. No spam, unsubscribe anytime.